2024-08-06T20:30:00.7227556Z Current runner version: '2.318.0' 2024-08-06T20:30:00.7237914Z Runner name: 'i-0de4eefb8fab684da' 2024-08-06T20:30:00.7239337Z Runner group name: 'Default' 2024-08-06T20:30:00.7240621Z Machine name: 'ip-10-0-62-112' 2024-08-06T20:30:00.7247087Z ##[group]GITHUB_TOKEN Permissions 2024-08-06T20:30:00.7250575Z Actions: read 2024-08-06T20:30:00.7251551Z Attestations: read 2024-08-06T20:30:00.7252393Z Checks: read 2024-08-06T20:30:00.7253311Z Contents: read 2024-08-06T20:30:00.7254290Z Deployments: read 2024-08-06T20:30:00.7255126Z Discussions: read 2024-08-06T20:30:00.7256030Z Issues: read 2024-08-06T20:30:00.7256862Z Metadata: read 2024-08-06T20:30:00.7257733Z Packages: read 2024-08-06T20:30:00.7258633Z Pages: read 2024-08-06T20:30:00.7259452Z PullRequests: read 2024-08-06T20:30:00.7260370Z RepositoryProjects: read 2024-08-06T20:30:00.7261375Z SecurityEvents: read 2024-08-06T20:30:00.7262260Z Statuses: read 2024-08-06T20:30:00.7263121Z ##[endgroup] 2024-08-06T20:30:00.7267985Z Secret source: Actions 2024-08-06T20:30:00.7269304Z Prepare workflow directory 2024-08-06T20:30:01.1691998Z Prepare all required actions 2024-08-06T20:30:01.1748451Z Getting action download info 2024-08-06T20:30:01.3613348Z Download action repository 'pytorch/test-infra@main' (SHA:a1f5a89251fc4258ab59806094fe3108f7d6741a) 2024-08-06T20:30:03.4009292Z Download action repository 'pytorch/pytorch@main' (SHA:de00c7958301ce81b9716bdef5731ed40d4d14ca) 2024-08-06T20:30:20.9873033Z Download action repository 'aws-actions/configure-aws-credentials@v3' (SHA:50ac8dd1e1b10d09dac7b8727528b91bed831ac0) 2024-08-06T20:30:21.2266675Z Download action repository 'seemethere/upload-artifact-s3@v5' (SHA:baba72d0712b404f646cebe0730933554ebce96a) 2024-08-06T20:30:21.5717027Z Getting action download info 2024-08-06T20:30:21.6622696Z Download action repository 'malfet/checkout@silent-checkout' (SHA:e07af140b3ccefc05679e3755b9db68f4ee4589c) 2024-08-06T20:30:22.0524193Z Getting action download info 2024-08-06T20:30:22.1432531Z Download action repository 'nick-fields/retry@3e91a01664abd3c5cd539100d10d33b9c5b68482' (SHA:3e91a01664abd3c5cd539100d10d33b9c5b68482) 2024-08-06T20:30:22.3666137Z Uses: pytorch/pytorch/.github/workflows/_linux-test.yml@refs/pull/132710/merge (bf5bb5a1585a03379137fab341e87c02c77e76cd) 2024-08-06T20:30:22.3668866Z ##[group] Inputs 2024-08-06T20:30:22.3669320Z build-environment: linux-focal-cuda12.1-py3.10-gcc9 2024-08-06T20:30:22.3671652Z test-matrix: {"include": [{"config": "default", "shard": 1, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 2, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 3, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 4, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 5, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}]} 2024-08-06T20:30:22.3674412Z docker-image: 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:30:22.3675397Z sync-tag: 2024-08-06T20:30:22.3676505Z timeout-minutes: 360 2024-08-06T20:30:22.3676869Z use-gha: 2024-08-06T20:30:22.3677152Z dashboard-tag: 2024-08-06T20:30:22.3677493Z s3-bucket: gha-artifacts 2024-08-06T20:30:22.3677865Z aws-role-to-assume: 2024-08-06T20:30:22.3678206Z ##[endgroup] 2024-08-06T20:30:22.3678834Z Complete job name: linux-focal-cuda12.1-py3.10-gcc9 / test (default, 1, 5, amz2023.linux.4xlarge.nvidia.gpu) 2024-08-06T20:30:22.4355392Z A job started hook has been configured by the self-hosted runner administrator 2024-08-06T20:30:22.4501164Z ##[group]Run '/home/ec2-user/runner-scripts/before_job.sh' 2024-08-06T20:30:22.4513065Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:30:22.4514136Z ##[endgroup] 2024-08-06T20:30:24.2597233Z Runner Type: amz2023.linux.4xlarge.nvidia.gpu 2024-08-06T20:30:24.2597887Z Instance Type: g3.4xlarge 2024-08-06T20:30:24.2598843Z AMI Name: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64 2024-08-06T20:30:24.2599382Z AMI ID: ami-06c68f701d8090592 2024-08-06T20:30:31.4911312Z ##[group]Run pytorch/test-infra/.github/actions/setup-ssh@main 2024-08-06T20:30:31.4911908Z with: 2024-08-06T20:30:31.4912700Z github-secret: *** 2024-08-06T20:30:31.4913610Z instructions: All testing is done inside the container, to start an interactive session run: docker exec -it $(docker container ps --format '{{.ID}}') bash 2024-08-06T20:30:31.4914559Z activate-with-label: false 2024-08-06T20:30:31.4914919Z label: with-ssh 2024-08-06T20:30:31.4915257Z remove-existing-keys: true 2024-08-06T20:30:31.4915633Z fail-silently: true 2024-08-06T20:30:31.4915952Z env: 2024-08-06T20:30:31.4916249Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:30:31.4916618Z ##[endgroup] 2024-08-06T20:30:31.6070877Z Please see https://github.com/pytorch/pytorch/wiki/Debugging-using-with-ssh-for-Github-Actions for more info. 2024-08-06T20:30:31.9228553Z Grabbing public ssh keys from https://github.com/drisspg.keys 2024-08-06T20:30:31.9946341Z ~/.ssh/authorized_keys file found on node, removing ~/.ssh and starting fresh 2024-08-06T20:30:31.9967733Z Public keys pulled and installed to /home/ec2-user/.ssh/authorized_keys 2024-08-06T20:30:32.0009930Z Login using: ssh ec2-user@ec2-52-205-255-249.compute-1.amazonaws.com 2024-08-06T20:30:32.0010683Z All testing is done inside the container, to start an interactive session run: 2024-08-06T20:30:32.0011385Z docker exec -it $(docker container ps --format '{{.ID}}') bash 2024-08-06T20:30:32.0156387Z ##[group]Run pytorch/pytorch/.github/actions/checkout-pytorch@main 2024-08-06T20:30:32.0156987Z with: 2024-08-06T20:30:32.0157276Z submodules: recursive 2024-08-06T20:30:32.0157626Z fetch-depth: 0 2024-08-06T20:30:32.0157941Z env: 2024-08-06T20:30:32.0158218Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:30:32.0158570Z ##[endgroup] 2024-08-06T20:30:32.0269513Z ##[group]Run retry () { 2024-08-06T20:30:32.0269912Z retry () { 2024-08-06T20:30:32.0270366Z  $* || (sleep 1 && $*) || (sleep 2 && $*) || (sleep 4 && $*) || (sleep 8 && $*) 2024-08-06T20:30:32.0270869Z } 2024-08-06T20:30:32.0271224Z echo "${GITHUB_WORKSPACE}" 2024-08-06T20:30:32.0271653Z if [ -z "${NO_SUDO}" ]; then 2024-08-06T20:30:32.0272105Z  retry sudo rm -rf "${GITHUB_WORKSPACE}" 2024-08-06T20:30:32.0272539Z else 2024-08-06T20:30:32.0272865Z  retry rm -rf "${GITHUB_WORKSPACE}" 2024-08-06T20:30:32.0273285Z fi 2024-08-06T20:30:32.0273651Z mkdir "${GITHUB_WORKSPACE}" 2024-08-06T20:30:32.0281522Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:30:32.0282022Z env: 2024-08-06T20:30:32.0282308Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:30:32.0282662Z NO_SUDO: 2024-08-06T20:30:32.0282958Z ##[endgroup] 2024-08-06T20:30:32.0314459Z /home/ec2-user/actions-runner/_work/pytorch/pytorch 2024-08-06T20:30:35.0990685Z ##[group]Run malfet/checkout@silent-checkout 2024-08-06T20:30:35.0991148Z with: 2024-08-06T20:30:35.0991461Z ref: b9d86fa89636e301796d4201f36d86c73f6e49bc 2024-08-06T20:30:35.0991891Z fetch-depth: 0 2024-08-06T20:30:35.0992211Z submodules: recursive 2024-08-06T20:30:35.0992544Z quiet-checkout: true 2024-08-06T20:30:35.0992897Z repository: pytorch/pytorch 2024-08-06T20:30:35.0993402Z token: *** 2024-08-06T20:30:35.0993702Z ssh-strict: true 2024-08-06T20:30:35.0994015Z persist-credentials: true 2024-08-06T20:30:35.0994372Z clean: true 2024-08-06T20:30:35.0994700Z sparse-checkout-cone-mode: true 2024-08-06T20:30:35.0995084Z lfs: false 2024-08-06T20:30:35.0995386Z set-safe-directory: true 2024-08-06T20:30:35.0995711Z env: 2024-08-06T20:30:35.0995994Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:30:35.0996334Z ##[endgroup] 2024-08-06T20:30:35.2337488Z Syncing repository: pytorch/pytorch 2024-08-06T20:30:35.2339335Z ##[group]Getting Git version info 2024-08-06T20:30:35.2339947Z Working directory is '/home/ec2-user/actions-runner/_work/pytorch/pytorch' 2024-08-06T20:30:35.2341183Z [command]/usr/bin/git version 2024-08-06T20:30:35.2341606Z git version 2.40.1 2024-08-06T20:30:35.2370583Z ##[endgroup] 2024-08-06T20:30:35.2391906Z Temporarily overriding HOME='/home/ec2-user/actions-runner/_work/_temp/b05a78bb-c54c-4684-9588-2df1faab20e3' before making global git config changes 2024-08-06T20:30:35.2393118Z Adding repository directory to the temporary git global config as a safe directory 2024-08-06T20:30:35.2399067Z [command]/usr/bin/git config --global --add safe.directory /home/ec2-user/actions-runner/_work/pytorch/pytorch 2024-08-06T20:30:35.2447440Z Deleting the contents of '/home/ec2-user/actions-runner/_work/pytorch/pytorch' 2024-08-06T20:30:35.2453458Z ##[group]Initializing the repository 2024-08-06T20:30:35.2457413Z [command]/usr/bin/git init /home/ec2-user/actions-runner/_work/pytorch/pytorch 2024-08-06T20:30:35.2495278Z hint: Using 'master' as the name for the initial branch. This default branch name 2024-08-06T20:30:35.2496166Z hint: is subject to change. To configure the initial branch name to use in all 2024-08-06T20:30:35.2496877Z hint: of your new repositories, which will suppress this warning, call: 2024-08-06T20:30:35.2497410Z hint: 2024-08-06T20:30:35.2497798Z hint: git config --global init.defaultBranch 2024-08-06T20:30:35.2498241Z hint: 2024-08-06T20:30:35.2498679Z hint: Names commonly chosen instead of 'master' are 'main', 'trunk' and 2024-08-06T20:30:35.2499390Z hint: 'development'. The just-created branch can be renamed via this command: 2024-08-06T20:30:35.2500347Z hint: 2024-08-06T20:30:35.2500739Z hint: git branch -m 2024-08-06T20:30:35.2501402Z Initialized empty Git repository in /home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/ 2024-08-06T20:30:35.2510922Z [command]/usr/bin/git remote add origin https://github.com/pytorch/pytorch 2024-08-06T20:30:35.2547079Z ##[endgroup] 2024-08-06T20:30:35.2547660Z ##[group]Disabling automatic garbage collection 2024-08-06T20:30:35.2550898Z [command]/usr/bin/git config --local gc.auto 0 2024-08-06T20:30:35.2586580Z ##[endgroup] 2024-08-06T20:30:35.2587098Z ##[group]Setting up auth 2024-08-06T20:30:35.2596147Z [command]/usr/bin/git config --local --name-only --get-regexp core\.sshCommand 2024-08-06T20:30:35.2633493Z [command]/usr/bin/git submodule foreach --recursive sh -c "git config --local --name-only --get-regexp 'core\.sshCommand' && git config --local --unset-all 'core.sshCommand' || :" 2024-08-06T20:30:35.2953594Z [command]/usr/bin/git config --local --name-only --get-regexp http\.https\:\/\/github\.com\/\.extraheader 2024-08-06T20:30:35.2987748Z [command]/usr/bin/git submodule foreach --recursive sh -c "git config --local --name-only --get-regexp 'http\.https\:\/\/github\.com\/\.extraheader' && git config --local --unset-all 'http.https://github.com/.extraheader' || :" 2024-08-06T20:30:35.3307366Z [command]/usr/bin/git config --local http.https://github.com/.extraheader AUTHORIZATION: basic *** 2024-08-06T20:30:35.3361193Z ##[endgroup] 2024-08-06T20:30:35.3362031Z ##[group]Fetching the repository 2024-08-06T20:30:35.3369843Z [command]/usr/bin/git -c protocol.version=2 fetch --prune --progress --no-recurse-submodules --quiet origin +refs/heads/*:refs/remotes/origin/* +refs/tags/*:refs/tags/* 2024-08-06T20:30:37.6788343Z remote: Enumerating objects: 1008367 2024-08-06T20:30:37.6788908Z remote: Enumerating objects: 1008693, done. 2024-08-06T20:30:37.6789413Z remote: Counting objects: 0% (1/326) 2024-08-06T20:30:37.6790011Z remote: Counting objects: 1% (4/326) 2024-08-06T20:30:37.6790520Z remote: Counting objects: 2% (7/326) 2024-08-06T20:30:37.6790994Z remote: Counting objects: 3% (10/326) 2024-08-06T20:30:37.6791453Z remote: Counting objects: 4% (14/326) 2024-08-06T20:30:37.6791922Z remote: Counting objects: 5% (17/326) 2024-08-06T20:30:37.6792399Z remote: Counting objects: 6% (20/326) 2024-08-06T20:30:37.6792857Z remote: Counting objects: 7% (23/326) 2024-08-06T20:30:37.6793750Z remote: Counting objects: 8% (27/326) 2024-08-06T20:30:37.6794226Z remote: Counting objects: 9% (30/326) 2024-08-06T20:30:37.6794678Z remote: Counting objects: 10% (33/326) 2024-08-06T20:30:37.6795147Z remote: Counting objects: 11% (36/326) 2024-08-06T20:30:37.6795687Z remote: Counting objects: 12% (40/326) 2024-08-06T20:30:37.6796161Z remote: Counting objects: 13% (43/326) 2024-08-06T20:30:37.6796614Z remote: Counting objects: 14% (46/326) 2024-08-06T20:30:37.6797101Z remote: Counting objects: 15% (49/326) 2024-08-06T20:30:37.6797582Z remote: Counting objects: 16% (53/326) 2024-08-06T20:30:37.6798036Z remote: Counting objects: 17% (56/326) 2024-08-06T20:30:37.6798511Z remote: Counting objects: 18% (59/326) 2024-08-06T20:30:37.6798984Z remote: Counting objects: 19% (62/326) 2024-08-06T20:30:37.6799439Z remote: Counting objects: 20% (66/326) 2024-08-06T20:30:37.6799914Z remote: Counting objects: 21% (69/326) 2024-08-06T20:30:37.6800398Z remote: Counting objects: 22% (72/326) 2024-08-06T20:30:37.6800855Z remote: Counting objects: 23% (75/326) 2024-08-06T20:30:37.6801328Z remote: Counting objects: 24% (79/326) 2024-08-06T20:30:37.6801801Z remote: Counting objects: 25% (82/326) 2024-08-06T20:30:37.6802254Z remote: Counting objects: 26% (85/326) 2024-08-06T20:30:37.6802836Z remote: Counting objects: 27% (89/326) 2024-08-06T20:30:37.6803510Z remote: Counting objects: 28% (92/326) 2024-08-06T20:30:37.6804452Z remote: Counting objects: 29% (95/326) 2024-08-06T20:30:37.6805049Z remote: Counting objects: 30% (98/326) 2024-08-06T20:30:37.6805527Z remote: Counting objects: 31% (102/326) 2024-08-06T20:30:37.6805988Z remote: Counting objects: 32% (105/326) 2024-08-06T20:30:37.6806466Z remote: Counting objects: 33% (108/326) 2024-08-06T20:30:37.6806951Z remote: Counting objects: 34% (111/326) 2024-08-06T20:30:37.6807440Z remote: Counting objects: 35% (115/326) 2024-08-06T20:30:37.6807894Z remote: Counting objects: 36% (118/326) 2024-08-06T20:30:37.6808371Z remote: Counting objects: 37% (121/326) 2024-08-06T20:30:37.6808855Z remote: Counting objects: 38% (124/326) 2024-08-06T20:30:37.6809316Z remote: Counting objects: 39% (128/326) 2024-08-06T20:30:37.6809793Z remote: Counting objects: 40% (131/326) 2024-08-06T20:30:37.6810275Z remote: Counting objects: 41% (134/326) 2024-08-06T20:30:37.6810731Z remote: Counting objects: 42% (137/326) 2024-08-06T20:30:37.6811205Z remote: Counting objects: 43% (141/326) 2024-08-06T20:30:37.6811676Z remote: Counting objects: 44% (144/326) 2024-08-06T20:30:37.6812134Z remote: Counting objects: 45% (147/326) 2024-08-06T20:30:37.6812611Z remote: Counting objects: 46% (150/326) 2024-08-06T20:30:37.6813088Z remote: Counting objects: 47% (154/326) 2024-08-06T20:30:37.6813553Z remote: Counting objects: 48% (157/326) 2024-08-06T20:30:37.6814027Z remote: Counting objects: 49% (160/326) 2024-08-06T20:30:37.6814501Z remote: Counting objects: 50% (163/326) 2024-08-06T20:30:37.6814963Z remote: Counting objects: 51% (167/326) 2024-08-06T20:30:37.6815438Z remote: Counting objects: 52% (170/326) 2024-08-06T20:30:37.6815915Z remote: Counting objects: 53% (173/326) 2024-08-06T20:30:37.6816443Z remote: Counting objects: 54% (177/326) 2024-08-06T20:30:37.6816925Z remote: Counting objects: 55% (180/326) 2024-08-06T20:30:37.6817401Z remote: Counting objects: 56% (183/326) 2024-08-06T20:30:37.6817856Z remote: Counting objects: 57% (186/326) 2024-08-06T20:30:37.6818328Z remote: Counting objects: 58% (190/326) 2024-08-06T20:30:37.6818804Z remote: Counting objects: 59% (193/326) 2024-08-06T20:30:37.6819255Z remote: Counting objects: 60% (196/326) 2024-08-06T20:30:37.6819909Z remote: Counting objects: 61% (199/326) 2024-08-06T20:30:37.6820382Z remote: Counting objects: 62% (203/326) 2024-08-06T20:30:37.6820897Z remote: Counting objects: 63% (206/326) 2024-08-06T20:30:37.6821360Z remote: Counting objects: 64% (209/326) 2024-08-06T20:30:37.6821838Z remote: Counting objects: 65% (212/326) 2024-08-06T20:30:37.6822309Z remote: Counting objects: 66% (216/326) 2024-08-06T20:30:37.6822765Z remote: Counting objects: 67% (219/326) 2024-08-06T20:30:37.6823247Z remote: Counting objects: 68% (222/326) 2024-08-06T20:30:37.6823725Z remote: Counting objects: 69% (225/326) 2024-08-06T20:30:37.6824180Z remote: Counting objects: 70% (229/326) 2024-08-06T20:30:37.6824700Z remote: Counting objects: 71% (232/326) 2024-08-06T20:30:37.6825211Z remote: Counting objects: 72% (235/326) 2024-08-06T20:30:37.6825669Z remote: Counting objects: 73% (238/326) 2024-08-06T20:30:37.6826172Z remote: Counting objects: 74% (242/326) 2024-08-06T20:30:37.6826644Z remote: Counting objects: 75% (245/326) 2024-08-06T20:30:37.6827097Z remote: Counting objects: 76% (248/326) 2024-08-06T20:30:37.6827566Z remote: Counting objects: 77% (252/326) 2024-08-06T20:30:37.6828037Z remote: Counting objects: 78% (255/326) 2024-08-06T20:30:37.6828493Z remote: Counting objects: 79% (258/326) 2024-08-06T20:30:37.6828970Z remote: Counting objects: 80% (261/326) 2024-08-06T20:30:37.6829567Z remote: Counting objects: 81% (265/326) 2024-08-06T20:30:37.6830047Z remote: Counting objects: 82% (268/326) 2024-08-06T20:30:37.6830526Z remote: Counting objects: 83% (271/326) 2024-08-06T20:30:37.6831000Z remote: Counting objects: 84% (274/326) 2024-08-06T20:30:37.6831460Z remote: Counting objects: 85% (278/326) 2024-08-06T20:30:37.6831936Z remote: Counting objects: 86% (281/326) 2024-08-06T20:30:37.6832420Z remote: Counting objects: 87% (284/326) 2024-08-06T20:30:37.6832886Z remote: Counting objects: 88% (287/326) 2024-08-06T20:30:37.6833359Z remote: Counting objects: 89% (291/326) 2024-08-06T20:30:37.6833832Z remote: Counting objects: 90% (294/326) 2024-08-06T20:30:37.6834288Z remote: Counting objects: 91% (297/326) 2024-08-06T20:30:37.6834762Z remote: Counting objects: 92% (300/326) 2024-08-06T20:30:37.6835239Z remote: Counting objects: 93% (304/326) 2024-08-06T20:30:37.6835716Z remote: Counting objects: 94% (307/326) 2024-08-06T20:30:37.6836196Z remote: Counting objects: 95% (310/326) 2024-08-06T20:30:37.6836671Z remote: Counting objects: 96% (313/326) 2024-08-06T20:30:37.6837148Z remote: Counting objects: 97% (317/326) 2024-08-06T20:30:37.6837605Z remote: Counting objects: 98% (320/326) 2024-08-06T20:30:37.6838081Z remote: Counting objects: 99% (323/326) 2024-08-06T20:30:37.6838557Z remote: Counting objects: 100% (326/326) 2024-08-06T20:30:37.6839057Z remote: Counting objects: 100% (326/326), done. 2024-08-06T20:30:37.6839576Z remote: Compressing objects: 0% (1/185) 2024-08-06T20:30:37.6874639Z remote: Compressing objects: 1% (2/185) 2024-08-06T20:30:37.6900939Z remote: Compressing objects: 2% (4/185) 2024-08-06T20:30:37.6926675Z remote: Compressing objects: 3% (6/185) 2024-08-06T20:30:37.7046919Z remote: Compressing objects: 4% (8/185) 2024-08-06T20:30:37.7177239Z remote: Compressing objects: 5% (10/185) 2024-08-06T20:30:37.7210565Z remote: Compressing objects: 6% (12/185) 2024-08-06T20:30:37.7239236Z remote: Compressing objects: 7% (13/185) 2024-08-06T20:30:37.7287665Z remote: Compressing objects: 8% (15/185) 2024-08-06T20:30:37.7298488Z remote: Compressing objects: 9% (17/185) 2024-08-06T20:30:37.7457193Z remote: Compressing objects: 10% (19/185) 2024-08-06T20:30:37.7458219Z remote: Compressing objects: 11% (21/185) 2024-08-06T20:30:37.7458695Z remote: Compressing objects: 12% (23/185) 2024-08-06T20:30:37.7459190Z remote: Compressing objects: 13% (25/185) 2024-08-06T20:30:37.7459680Z remote: Compressing objects: 14% (26/185) 2024-08-06T20:30:37.7460151Z remote: Compressing objects: 15% (28/185) 2024-08-06T20:30:37.7460634Z remote: Compressing objects: 16% (30/185) 2024-08-06T20:30:37.7461124Z remote: Compressing objects: 17% (32/185) 2024-08-06T20:30:37.7461603Z remote: Compressing objects: 18% (34/185) 2024-08-06T20:30:37.7462095Z remote: Compressing objects: 19% (36/185) 2024-08-06T20:30:37.7462579Z remote: Compressing objects: 20% (37/185) 2024-08-06T20:30:37.7463044Z remote: Compressing objects: 21% (39/185) 2024-08-06T20:30:37.7463529Z remote: Compressing objects: 22% (41/185) 2024-08-06T20:30:37.7464011Z remote: Compressing objects: 23% (43/185) 2024-08-06T20:30:37.7464488Z remote: Compressing objects: 24% (45/185) 2024-08-06T20:30:37.7464973Z remote: Compressing objects: 25% (47/185) 2024-08-06T20:30:37.7465482Z remote: Compressing objects: 26% (49/185) 2024-08-06T20:30:37.7465965Z remote: Compressing objects: 27% (50/185) 2024-08-06T20:30:37.7466430Z remote: Compressing objects: 28% (52/185) 2024-08-06T20:30:37.7466922Z remote: Compressing objects: 29% (54/185) 2024-08-06T20:30:37.7468923Z remote: Compressing objects: 30% (56/185) 2024-08-06T20:30:37.7469464Z remote: Compressing objects: 31% (58/185) 2024-08-06T20:30:37.7470430Z remote: Compressing objects: 32% (60/185) 2024-08-06T20:30:37.7473679Z remote: Compressing objects: 33% (62/185) 2024-08-06T20:30:37.7477842Z remote: Compressing objects: 34% (63/185) 2024-08-06T20:30:37.7484418Z remote: Compressing objects: 35% (65/185) 2024-08-06T20:30:37.7486713Z remote: Compressing objects: 36% (67/185) 2024-08-06T20:30:37.7487219Z remote: Compressing objects: 37% (69/185) 2024-08-06T20:30:37.7496576Z remote: Compressing objects: 38% (71/185) 2024-08-06T20:30:37.7497593Z remote: Compressing objects: 39% (73/185) 2024-08-06T20:30:37.7498063Z remote: Compressing objects: 40% (74/185) 2024-08-06T20:30:37.7498574Z remote: Compressing objects: 41% (76/185) 2024-08-06T20:30:37.7499060Z remote: Compressing objects: 42% (78/185) 2024-08-06T20:30:37.7499572Z remote: Compressing objects: 43% (80/185) 2024-08-06T20:30:37.7525255Z remote: Compressing objects: 44% (82/185) 2024-08-06T20:30:37.7525789Z remote: Compressing objects: 45% (84/185) 2024-08-06T20:30:37.7526281Z remote: Compressing objects: 46% (86/185) 2024-08-06T20:30:37.7526760Z remote: Compressing objects: 47% (87/185) 2024-08-06T20:30:37.7527257Z remote: Compressing objects: 48% (89/185) 2024-08-06T20:30:37.7527752Z remote: Compressing objects: 49% (91/185) 2024-08-06T20:30:37.7528231Z remote: Compressing objects: 50% (93/185) 2024-08-06T20:30:37.7528727Z remote: Compressing objects: 51% (95/185) 2024-08-06T20:30:37.7529219Z remote: Compressing objects: 52% (97/185) 2024-08-06T20:30:37.7529692Z remote: Compressing objects: 53% (99/185) 2024-08-06T20:30:37.7530186Z remote: Compressing objects: 54% (100/185) 2024-08-06T20:30:37.7530686Z remote: Compressing objects: 55% (102/185) 2024-08-06T20:30:37.7531170Z remote: Compressing objects: 56% (104/185) 2024-08-06T20:30:37.7531670Z remote: Compressing objects: 57% (106/185) 2024-08-06T20:30:37.7532166Z remote: Compressing objects: 58% (108/185) 2024-08-06T20:30:37.7532645Z remote: Compressing objects: 59% (110/185) 2024-08-06T20:30:37.7533140Z remote: Compressing objects: 60% (111/185) 2024-08-06T20:30:37.7533632Z remote: Compressing objects: 61% (113/185) 2024-08-06T20:30:37.7534106Z remote: Compressing objects: 62% (115/185) 2024-08-06T20:30:37.7534970Z remote: Compressing objects: 63% (117/185) 2024-08-06T20:30:37.7535469Z remote: Compressing objects: 64% (119/185) 2024-08-06T20:30:37.7535966Z remote: Compressing objects: 65% (121/185) 2024-08-06T20:30:37.7536442Z remote: Compressing objects: 66% (123/185) 2024-08-06T20:30:37.7536931Z remote: Compressing objects: 67% (124/185) 2024-08-06T20:30:37.7537422Z remote: Compressing objects: 68% (126/185) 2024-08-06T20:30:37.7537904Z remote: Compressing objects: 69% (128/185) 2024-08-06T20:30:37.7538399Z remote: Compressing objects: 70% (130/185) 2024-08-06T20:30:37.7538888Z remote: Compressing objects: 71% (132/185) 2024-08-06T20:30:37.7539357Z remote: Compressing objects: 72% (134/185) 2024-08-06T20:30:37.7539845Z remote: Compressing objects: 73% (136/185) 2024-08-06T20:30:37.7540338Z remote: Compressing objects: 74% (137/185) 2024-08-06T20:30:37.7540824Z remote: Compressing objects: 75% (139/185) 2024-08-06T20:30:37.7541314Z remote: Compressing objects: 76% (141/185) 2024-08-06T20:30:37.7541799Z remote: Compressing objects: 77% (143/185) 2024-08-06T20:30:37.7542271Z remote: Compressing objects: 78% (145/185) 2024-08-06T20:30:37.7542761Z remote: Compressing objects: 79% (147/185) 2024-08-06T20:30:37.7543264Z remote: Compressing objects: 80% (148/185) 2024-08-06T20:30:37.7543868Z remote: Compressing objects: 81% (150/185) 2024-08-06T20:30:37.7544386Z remote: Compressing objects: 82% (152/185) 2024-08-06T20:30:37.7544878Z remote: Compressing objects: 83% (154/185) 2024-08-06T20:30:37.7545352Z remote: Compressing objects: 84% (156/185) 2024-08-06T20:30:37.7545838Z remote: Compressing objects: 85% (158/185) 2024-08-06T20:30:37.7546327Z remote: Compressing objects: 86% (160/185) 2024-08-06T20:30:37.7546796Z remote: Compressing objects: 87% (161/185) 2024-08-06T20:30:37.7547302Z remote: Compressing objects: 88% (163/185) 2024-08-06T20:30:37.7547797Z remote: Compressing objects: 89% (165/185) 2024-08-06T20:30:37.7548271Z remote: Compressing objects: 90% (167/185) 2024-08-06T20:30:37.7548770Z remote: Compressing objects: 91% (169/185) 2024-08-06T20:30:37.7549264Z remote: Compressing objects: 92% (171/185) 2024-08-06T20:30:37.7549739Z remote: Compressing objects: 93% (173/185) 2024-08-06T20:30:37.7550249Z remote: Compressing objects: 94% (174/185) 2024-08-06T20:30:37.7550749Z remote: Compressing objects: 95% (176/185) 2024-08-06T20:30:37.7551224Z remote: Compressing objects: 96% (178/185) 2024-08-06T20:30:37.7551723Z remote: Compressing objects: 97% (180/185) 2024-08-06T20:30:37.7552223Z remote: Compressing objects: 98% (182/185) 2024-08-06T20:30:37.7552718Z remote: Compressing objects: 99% (184/185) 2024-08-06T20:30:37.7553209Z remote: Compressing objects: 100% (185/185) 2024-08-06T20:30:37.7553745Z remote: Compressing objects: 100% (185/185), done. 2024-08-06T20:31:04.8813289Z remote: Total 1008693 (delta 160), reused 268 (delta 139), pack-reused 1008367 2024-08-06T20:31:30.9282528Z [command]/usr/bin/git rev-parse --verify --quiet b9d86fa89636e301796d4201f36d86c73f6e49bc^{object} 2024-08-06T20:31:30.9314423Z b9d86fa89636e301796d4201f36d86c73f6e49bc 2024-08-06T20:31:30.9320089Z ##[endgroup] 2024-08-06T20:31:30.9320706Z ##[group]Determining the checkout info 2024-08-06T20:31:30.9321900Z ##[endgroup] 2024-08-06T20:31:30.9322423Z ##[group]Checking out the ref 2024-08-06T20:31:30.9326437Z [command]/usr/bin/git checkout --quiet --force b9d86fa89636e301796d4201f36d86c73f6e49bc 2024-08-06T20:31:32.8156808Z ##[endgroup] 2024-08-06T20:31:32.8157400Z ##[group]Setting up auth for fetching submodules 2024-08-06T20:31:32.8163096Z [command]/usr/bin/git config --global http.https://github.com/.extraheader AUTHORIZATION: basic *** 2024-08-06T20:31:32.8226169Z [command]/usr/bin/git config --global --unset-all url.https://github.com/.insteadOf 2024-08-06T20:31:32.8262589Z [command]/usr/bin/git config --global --add url.https://github.com/.insteadOf git@github.com: 2024-08-06T20:31:32.8299545Z [command]/usr/bin/git config --global --add url.https://github.com/.insteadOf org-21003710@github.com: 2024-08-06T20:31:32.8332008Z ##[endgroup] 2024-08-06T20:31:32.8332522Z ##[group]Fetching submodules 2024-08-06T20:31:32.8337404Z [command]/usr/bin/git submodule sync --recursive 2024-08-06T20:31:32.8688989Z [command]/usr/bin/git -c protocol.version=2 submodule update --init --force --recursive 2024-08-06T20:31:32.9026039Z Submodule 'android/libs/fbjni' (https://github.com/facebookincubator/fbjni.git) registered for path 'android/libs/fbjni' 2024-08-06T20:31:32.9028936Z Submodule 'third_party/NNPACK_deps/FP16' (https://github.com/Maratyszcza/FP16.git) registered for path 'third_party/FP16' 2024-08-06T20:31:32.9033472Z Submodule 'third_party/NNPACK_deps/FXdiv' (https://github.com/Maratyszcza/FXdiv.git) registered for path 'third_party/FXdiv' 2024-08-06T20:31:32.9037178Z Submodule 'third_party/NNPACK' (https://github.com/Maratyszcza/NNPACK.git) registered for path 'third_party/NNPACK' 2024-08-06T20:31:32.9041467Z Submodule 'third_party/VulkanMemoryAllocator' (https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator.git) registered for path 'third_party/VulkanMemoryAllocator' 2024-08-06T20:31:32.9045334Z Submodule 'third_party/XNNPACK' (https://github.com/google/XNNPACK.git) registered for path 'third_party/XNNPACK' 2024-08-06T20:31:32.9049336Z Submodule 'third_party/benchmark' (https://github.com/google/benchmark.git) registered for path 'third_party/benchmark' 2024-08-06T20:31:32.9053624Z Submodule 'third_party/cpp-httplib' (https://github.com/yhirose/cpp-httplib.git) registered for path 'third_party/cpp-httplib' 2024-08-06T20:31:32.9058028Z Submodule 'third_party/cpuinfo' (https://github.com/pytorch/cpuinfo.git) registered for path 'third_party/cpuinfo' 2024-08-06T20:31:32.9063219Z Submodule 'third_party/cudnn_frontend' (https://github.com/NVIDIA/cudnn-frontend.git) registered for path 'third_party/cudnn_frontend' 2024-08-06T20:31:32.9067795Z Submodule 'third_party/cutlass' (https://github.com/NVIDIA/cutlass.git) registered for path 'third_party/cutlass' 2024-08-06T20:31:32.9072496Z Submodule 'third_party/eigen' (https://gitlab.com/libeigen/eigen.git) registered for path 'third_party/eigen' 2024-08-06T20:31:32.9077462Z Submodule 'third_party/fbgemm' (https://github.com/pytorch/fbgemm) registered for path 'third_party/fbgemm' 2024-08-06T20:31:32.9082529Z Submodule 'third_party/flatbuffers' (https://github.com/google/flatbuffers.git) registered for path 'third_party/flatbuffers' 2024-08-06T20:31:32.9087565Z Submodule 'third_party/fmt' (https://github.com/fmtlib/fmt.git) registered for path 'third_party/fmt' 2024-08-06T20:31:32.9092756Z Submodule 'third_party/foxi' (https://github.com/houseroad/foxi.git) registered for path 'third_party/foxi' 2024-08-06T20:31:32.9100191Z Submodule 'third_party/gemmlowp/gemmlowp' (https://github.com/google/gemmlowp.git) registered for path 'third_party/gemmlowp/gemmlowp' 2024-08-06T20:31:32.9106041Z Submodule 'third_party/gloo' (https://github.com/facebookincubator/gloo) registered for path 'third_party/gloo' 2024-08-06T20:31:32.9111508Z Submodule 'third_party/googletest' (https://github.com/google/googletest.git) registered for path 'third_party/googletest' 2024-08-06T20:31:32.9117370Z Submodule 'third_party/ideep' (https://github.com/intel/ideep) registered for path 'third_party/ideep' 2024-08-06T20:31:32.9123124Z Submodule 'third_party/ittapi' (https://github.com/intel/ittapi.git) registered for path 'third_party/ittapi' 2024-08-06T20:31:32.9129038Z Submodule 'third_party/kineto' (https://github.com/pytorch/kineto) registered for path 'third_party/kineto' 2024-08-06T20:31:32.9135253Z Submodule 'third_party/mimalloc' (https://github.com/microsoft/mimalloc.git) registered for path 'third_party/mimalloc' 2024-08-06T20:31:32.9141275Z Submodule 'third_party/nccl/nccl' (https://github.com/NVIDIA/nccl) registered for path 'third_party/nccl/nccl' 2024-08-06T20:31:32.9147502Z Submodule 'third_party/nlohmann' (https://github.com/nlohmann/json.git) registered for path 'third_party/nlohmann' 2024-08-06T20:31:32.9153856Z Submodule 'third_party/onnx' (https://github.com/onnx/onnx.git) registered for path 'third_party/onnx' 2024-08-06T20:31:32.9160783Z Submodule 'third_party/opentelemetry-cpp' (https://github.com/open-telemetry/opentelemetry-cpp.git) registered for path 'third_party/opentelemetry-cpp' 2024-08-06T20:31:32.9167245Z Submodule 'third_party/pocketfft' (https://github.com/mreineck/pocketfft) registered for path 'third_party/pocketfft' 2024-08-06T20:31:32.9174049Z Submodule 'third_party/protobuf' (https://github.com/protocolbuffers/protobuf.git) registered for path 'third_party/protobuf' 2024-08-06T20:31:32.9180961Z Submodule 'third_party/NNPACK_deps/psimd' (https://github.com/Maratyszcza/psimd.git) registered for path 'third_party/psimd' 2024-08-06T20:31:32.9188393Z Submodule 'third_party/NNPACK_deps/pthreadpool' (https://github.com/Maratyszcza/pthreadpool.git) registered for path 'third_party/pthreadpool' 2024-08-06T20:31:32.9195360Z Submodule 'third_party/pybind11' (https://github.com/pybind/pybind11.git) registered for path 'third_party/pybind11' 2024-08-06T20:31:32.9202619Z Submodule 'third_party/python-peachpy' (https://github.com/malfet/PeachPy.git) registered for path 'third_party/python-peachpy' 2024-08-06T20:31:32.9212508Z Submodule 'third_party/sleef' (https://github.com/shibatch/sleef) registered for path 'third_party/sleef' 2024-08-06T20:31:32.9219849Z Submodule 'third_party/tensorpipe' (https://github.com/pytorch/tensorpipe.git) registered for path 'third_party/tensorpipe' 2024-08-06T20:31:32.9251513Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/android/libs/fbjni'... 2024-08-06T20:31:33.2986595Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/FP16'... 2024-08-06T20:31:33.4944674Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/FXdiv'... 2024-08-06T20:31:33.7176668Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/NNPACK'... 2024-08-06T20:31:33.9789009Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/VulkanMemoryAllocator'... 2024-08-06T20:31:36.1940249Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/XNNPACK'... 2024-08-06T20:31:49.5226021Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/benchmark'... 2024-08-06T20:31:50.1109237Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/cpp-httplib'... 2024-08-06T20:31:50.6052662Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/cpuinfo'... 2024-08-06T20:31:51.2895509Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/cudnn_frontend'... 2024-08-06T20:31:52.7486083Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/cutlass'... 2024-08-06T20:31:55.1135958Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/eigen'... 2024-08-06T20:32:02.2854720Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/fbgemm'... 2024-08-06T20:32:03.9472582Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/flatbuffers'... 2024-08-06T20:32:05.9563680Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/fmt'... 2024-08-06T20:32:07.5620610Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/foxi'... 2024-08-06T20:32:07.7491385Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/gemmlowp/gemmlowp'... 2024-08-06T20:32:08.2451209Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/gloo'... 2024-08-06T20:32:08.6029077Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/googletest'... 2024-08-06T20:32:09.8584795Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/ideep'... 2024-08-06T20:32:10.2483770Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/ittapi'... 2024-08-06T20:32:10.5321810Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/kineto'... 2024-08-06T20:32:12.4513401Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/mimalloc'... 2024-08-06T20:32:13.2954091Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/nccl/nccl'... 2024-08-06T20:32:14.0323474Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/nlohmann'... 2024-08-06T20:32:21.6998853Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/onnx'... 2024-08-06T20:32:24.1031781Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/opentelemetry-cpp'... 2024-08-06T20:32:29.2391058Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/pocketfft'... 2024-08-06T20:32:29.4701389Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/protobuf'... 2024-08-06T20:32:39.9631120Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/psimd'... 2024-08-06T20:32:40.1393285Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/pthreadpool'... 2024-08-06T20:32:40.3498746Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/pybind11'... 2024-08-06T20:32:41.3624902Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/python-peachpy'... 2024-08-06T20:32:41.6886646Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/sleef'... 2024-08-06T20:32:42.3645291Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/tensorpipe'... 2024-08-06T20:32:42.8479102Z Submodule path 'android/libs/fbjni': checked out '7e1e1fe3858c63c251c637ae41a20de425dde96f' 2024-08-06T20:32:42.8625721Z Submodule path 'third_party/FP16': checked out '4dfe081cf6bcd15db339cf2680b9281b8451eeb3' 2024-08-06T20:32:42.8737817Z Submodule path 'third_party/FXdiv': checked out 'b408327ac2a15ec3e43352421954f5b1967701d1' 2024-08-06T20:32:42.9057335Z Submodule path 'third_party/NNPACK': checked out 'c07e3a0400713d546e0dea2d5466dd22ea389c73' 2024-08-06T20:32:42.9545674Z Submodule path 'third_party/VulkanMemoryAllocator': checked out 'a6bfc237255a6bac1513f7c1ebde6d8aed6b5191' 2024-08-06T20:32:44.2643929Z Submodule path 'third_party/XNNPACK': checked out 'fcbf55af6cf28a4627bcd1f703ab7ad843f0f3a2' 2024-08-06T20:32:44.2942558Z Submodule path 'third_party/benchmark': checked out '0d98dba29d66e93259db7daa53a9327df767a415' 2024-08-06T20:32:44.3493448Z Submodule path 'third_party/cpp-httplib': checked out '3b6597bba913d51161383657829b7e644e59c006' 2024-08-06T20:32:44.4758556Z Submodule path 'third_party/cpuinfo': checked out '3c8b1533ac03dd6531ab6e7b9245d488f13a82a5' 2024-08-06T20:32:44.5197401Z Submodule path 'third_party/cudnn_frontend': checked out '98ca4e1941fe3263f128f74f10063a3ea35c7019' 2024-08-06T20:32:45.2028911Z Submodule path 'third_party/cutlass': checked out 'bbe579a9e3beb6ea6626d9227ec32d0dae119a49' 2024-08-06T20:32:45.5373564Z Submodule path 'third_party/eigen': checked out '3147391d946bb4b6c68edd901f2add6ac1f31f8c' 2024-08-06T20:32:45.6405135Z Submodule path 'third_party/fbgemm': checked out 'dbc3157bf256f1339b3fa1fef2be89ac4078be0e' 2024-08-06T20:32:45.6425845Z Submodule 'third_party/asmjit' (https://github.com/asmjit/asmjit.git) registered for path 'third_party/fbgemm/third_party/asmjit' 2024-08-06T20:32:45.6429780Z Submodule 'third_party/cpuinfo' (https://github.com/pytorch/cpuinfo) registered for path 'third_party/fbgemm/third_party/cpuinfo' 2024-08-06T20:32:45.6433538Z Submodule 'third_party/cutlass' (https://github.com/NVIDIA/cutlass.git) registered for path 'third_party/fbgemm/third_party/cutlass' 2024-08-06T20:32:45.6437439Z Submodule 'third_party/googletest' (https://github.com/google/googletest) registered for path 'third_party/fbgemm/third_party/googletest' 2024-08-06T20:32:45.6441702Z Submodule 'third_party/hipify_torch' (https://github.com/ROCmSoftwarePlatform/hipify_torch.git) registered for path 'third_party/fbgemm/third_party/hipify_torch' 2024-08-06T20:32:45.6473234Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/fbgemm/third_party/asmjit'... 2024-08-06T20:32:46.8120889Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/fbgemm/third_party/cpuinfo'... 2024-08-06T20:32:47.4905927Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/fbgemm/third_party/cutlass'... 2024-08-06T20:32:49.8695487Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/fbgemm/third_party/googletest'... 2024-08-06T20:32:51.1683415Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/fbgemm/third_party/hipify_torch'... 2024-08-06T20:32:51.5890435Z Submodule path 'third_party/fbgemm/third_party/asmjit': checked out 'd3fbf7c9bc7c1d1365a94a45614b91c5a3706b81' 2024-08-06T20:32:51.7121597Z Submodule path 'third_party/fbgemm/third_party/cpuinfo': checked out 'ed8b86a253800bafdb7b25c5c399f91bff9cb1f3' 2024-08-06T20:32:52.2690191Z Submodule path 'third_party/fbgemm/third_party/cutlass': checked out 'fc9ebc645b63f3a6bc80aaefde5c063fb72110d6' 2024-08-06T20:32:52.3487178Z Submodule path 'third_party/fbgemm/third_party/googletest': checked out 'cbf019de22c8dd37b2108da35b2748fd702d1796' 2024-08-06T20:32:52.3635434Z Submodule path 'third_party/fbgemm/third_party/hipify_torch': checked out '23f53b025b466d8ec3c45d52290d3442f7fbe6b1' 2024-08-06T20:32:52.5224559Z Submodule path 'third_party/flatbuffers': checked out '01834de25e4bf3975a9a00e816292b1ad0fe184b' 2024-08-06T20:32:52.5727655Z Submodule path 'third_party/fmt': checked out '0c9fce2ffefecfdce794e1859584e25877b7b592' 2024-08-06T20:32:52.5843134Z Submodule path 'third_party/foxi': checked out 'c278588e34e535f0bb8f00df3880d26928038cad' 2024-08-06T20:32:52.6320776Z Submodule path 'third_party/gemmlowp/gemmlowp': checked out '3fb5c176c17c765a3492cd2f0321b0dab712f350' 2024-08-06T20:32:52.6651594Z Submodule path 'third_party/gloo': checked out '5354032ea08eadd7fc4456477f7f7c6308818509' 2024-08-06T20:32:52.7220709Z Submodule path 'third_party/googletest': checked out 'e2239ee6043f73722e7aa812a459f54a28552929' 2024-08-06T20:32:52.7376715Z Submodule path 'third_party/ideep': checked out '55ca0191687aaf19aca5cdb7881c791e3bea442b' 2024-08-06T20:32:52.7396642Z Submodule 'mkl-dnn' (https://github.com/intel/mkl-dnn.git) registered for path 'third_party/ideep/mkl-dnn' 2024-08-06T20:32:52.7425663Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/ideep/mkl-dnn'... 2024-08-06T20:33:08.6043752Z Submodule path 'third_party/ideep/mkl-dnn': checked out '1137e04ec0b5251ca2b4400a4fd3c667ce843d67' 2024-08-06T20:33:08.6267468Z Submodule path 'third_party/ittapi': checked out '5b8a7d7422611c3a0d799fb5fc5dd4abfae35b42' 2024-08-06T20:33:08.7362691Z Submodule path 'third_party/kineto': checked out 'da2f2682cabaf95d601fa2a9b7e0979f84fe7667' 2024-08-06T20:33:08.7383557Z Submodule 'libkineto/third_party/dynolog' (https://github.com/facebookincubator/dynolog.git) registered for path 'third_party/kineto/libkineto/third_party/dynolog' 2024-08-06T20:33:08.7387248Z Submodule 'libkineto/third_party/fmt' (https://github.com/fmtlib/fmt.git) registered for path 'third_party/kineto/libkineto/third_party/fmt' 2024-08-06T20:33:08.7391265Z Submodule 'libkineto/third_party/googletest' (https://github.com/google/googletest.git) registered for path 'third_party/kineto/libkineto/third_party/googletest' 2024-08-06T20:33:08.7423437Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/kineto/libkineto/third_party/dynolog'... 2024-08-06T20:33:10.0409328Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/kineto/libkineto/third_party/fmt'... 2024-08-06T20:33:11.4673525Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/kineto/libkineto/third_party/googletest'... 2024-08-06T20:33:12.8297792Z Submodule path 'third_party/kineto/libkineto/third_party/dynolog': checked out '7d04a0053a845370ae06ce317a22a48e9edcc74e' 2024-08-06T20:33:12.8318196Z Submodule 'third_party/DCGM' (https://github.com/NVIDIA/DCGM.git) registered for path 'third_party/kineto/libkineto/third_party/dynolog/third_party/DCGM' 2024-08-06T20:33:12.8321809Z Submodule 'third_party/cpr' (https://github.com/libcpr/cpr.git) registered for path 'third_party/kineto/libkineto/third_party/dynolog/third_party/cpr' 2024-08-06T20:33:12.8325669Z Submodule 'third_party/fmt' (https://github.com/fmtlib/fmt.git) registered for path 'third_party/kineto/libkineto/third_party/dynolog/third_party/fmt' 2024-08-06T20:33:12.8329573Z Submodule 'third_party/gflags' (https://github.com/gflags/gflags.git) registered for path 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags' 2024-08-06T20:33:12.8333614Z Submodule 'third_party/glog' (https://github.com/google/glog.git) registered for path 'third_party/kineto/libkineto/third_party/dynolog/third_party/glog' 2024-08-06T20:33:12.8337843Z Submodule 'third_party/googletest' (https://github.com/google/googletest.git) registered for path 'third_party/kineto/libkineto/third_party/dynolog/third_party/googletest' 2024-08-06T20:33:12.8342141Z Submodule 'third_party/json' (https://github.com/nlohmann/json.git) registered for path 'third_party/kineto/libkineto/third_party/dynolog/third_party/json' 2024-08-06T20:33:12.8346679Z Submodule 'third_party/pfs' (https://github.com/dtrugman/pfs.git) registered for path 'third_party/kineto/libkineto/third_party/dynolog/third_party/pfs' 2024-08-06T20:33:12.8378187Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/kineto/libkineto/third_party/dynolog/third_party/DCGM'... 2024-08-06T20:33:13.9119118Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/kineto/libkineto/third_party/dynolog/third_party/cpr'... 2024-08-06T20:33:14.3316330Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/kineto/libkineto/third_party/dynolog/third_party/fmt'... 2024-08-06T20:33:15.7312766Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/kineto/libkineto/third_party/dynolog/third_party/gflags'... 2024-08-06T20:33:16.0963168Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/kineto/libkineto/third_party/dynolog/third_party/glog'... 2024-08-06T20:33:16.7196800Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/kineto/libkineto/third_party/dynolog/third_party/googletest'... 2024-08-06T20:33:18.0451407Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/kineto/libkineto/third_party/dynolog/third_party/json'... 2024-08-06T20:33:25.7669428Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/kineto/libkineto/third_party/dynolog/third_party/pfs'... 2024-08-06T20:33:26.3362786Z Submodule path 'third_party/kineto/libkineto/third_party/dynolog/third_party/DCGM': checked out 'ffde4e54bc7249a6039a5e6b45b395141e1217f9' 2024-08-06T20:33:26.3601079Z Submodule path 'third_party/kineto/libkineto/third_party/dynolog/third_party/cpr': checked out '871ed52d350214a034f6ef8a3b8f51c5ce1bd400' 2024-08-06T20:33:26.4069150Z Submodule path 'third_party/kineto/libkineto/third_party/dynolog/third_party/fmt': checked out 'cd4af11efc9c622896a3e4cb599fa28668ca3d05' 2024-08-06T20:33:26.4235429Z Submodule path 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags': checked out 'e171aa2d15ed9eb17054558e0b3a6a413bb01067' 2024-08-06T20:33:26.4254604Z Submodule 'doc' (https://github.com/gflags/gflags.git) registered for path 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags/doc' 2024-08-06T20:33:26.4284995Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/kineto/libkineto/third_party/dynolog/third_party/gflags/doc'... 2024-08-06T20:33:26.7665744Z Submodule path 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags/doc': checked out '8411df715cf522606e3b1aca386ddfc0b63d34b4' 2024-08-06T20:33:26.7901852Z Submodule path 'third_party/kineto/libkineto/third_party/dynolog/third_party/glog': checked out 'b33e3bad4c46c8a6345525fd822af355e5ef9446' 2024-08-06T20:33:26.8423827Z Submodule path 'third_party/kineto/libkineto/third_party/dynolog/third_party/googletest': checked out '58d77fa8070e8cec2dc1ed015d66b454c8d78850' 2024-08-06T20:33:26.9725658Z Submodule path 'third_party/kineto/libkineto/third_party/dynolog/third_party/json': checked out '4f8fba14066156b73f1189a2b8bd568bde5284c5' 2024-08-06T20:33:26.9936674Z Submodule path 'third_party/kineto/libkineto/third_party/dynolog/third_party/pfs': checked out 'f68a2fa8ea36c783bdd760371411fcb495aa3150' 2024-08-06T20:33:27.0416074Z Submodule path 'third_party/kineto/libkineto/third_party/fmt': checked out '0041a40c1350ba702d475b9c4ad62da77caea164' 2024-08-06T20:33:27.1137774Z Submodule path 'third_party/kineto/libkineto/third_party/googletest': checked out '7aca84427f224eeed3144123d5230d5871e93347' 2024-08-06T20:33:27.1614916Z Submodule path 'third_party/mimalloc': checked out 'b66e3214d8a104669c2ec05ae91ebc26a8f5ab78' 2024-08-06T20:33:27.1923680Z Submodule path 'third_party/nccl/nccl': checked out 'ab2b89c4c339bd7f816fbc114a4b05d386b66290' 2024-08-06T20:33:27.3270042Z Submodule path 'third_party/nlohmann': checked out '87cda1d6646592ac5866dc703c8e1839046a6806' 2024-08-06T20:33:27.8372930Z Submodule path 'third_party/onnx': checked out '3bf92c03a9f27eba3bda1e5b9e63ea20ec213557' 2024-08-06T20:33:27.8412915Z Submodule 'third_party/benchmark' (https://github.com/google/benchmark.git) registered for path 'third_party/onnx/third_party/benchmark' 2024-08-06T20:33:27.8416466Z Submodule 'third_party/pybind11' (https://github.com/pybind/pybind11.git) registered for path 'third_party/onnx/third_party/pybind11' 2024-08-06T20:33:27.8447501Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/onnx/third_party/benchmark'... 2024-08-06T20:33:28.3598782Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/onnx/third_party/pybind11'... 2024-08-06T20:33:29.4347676Z Submodule path 'third_party/onnx/third_party/benchmark': checked out '2dd015dfef425c866d9a43f2c67d8b52d709acb6' 2024-08-06T20:33:29.4787908Z Submodule path 'third_party/onnx/third_party/pybind11': checked out '5b0a6fc2017fcc176545afe3e09c9f9885283242' 2024-08-06T20:33:29.5731568Z Submodule path 'third_party/opentelemetry-cpp': checked out 'a799f4aed9c94b765dcdaabaeab7d5e7e2310878' 2024-08-06T20:33:29.5752884Z Submodule 'third_party/benchmark' (https://github.com/google/benchmark) registered for path 'third_party/opentelemetry-cpp/third_party/benchmark' 2024-08-06T20:33:29.5756627Z Submodule 'third_party/googletest' (https://github.com/google/googletest) registered for path 'third_party/opentelemetry-cpp/third_party/googletest' 2024-08-06T20:33:29.5760267Z Submodule 'third_party/ms-gsl' (https://github.com/microsoft/GSL) registered for path 'third_party/opentelemetry-cpp/third_party/ms-gsl' 2024-08-06T20:33:29.5764178Z Submodule 'third_party/nlohmann-json' (https://github.com/nlohmann/json) registered for path 'third_party/opentelemetry-cpp/third_party/nlohmann-json' 2024-08-06T20:33:29.5768242Z Submodule 'third_party/opentelemetry-proto' (https://github.com/open-telemetry/opentelemetry-proto) registered for path 'third_party/opentelemetry-cpp/third_party/opentelemetry-proto' 2024-08-06T20:33:29.5772304Z Submodule 'third_party/opentracing-cpp' (https://github.com/opentracing/opentracing-cpp.git) registered for path 'third_party/opentelemetry-cpp/third_party/opentracing-cpp' 2024-08-06T20:33:29.5776388Z Submodule 'third_party/prometheus-cpp' (https://github.com/jupp0r/prometheus-cpp) registered for path 'third_party/opentelemetry-cpp/third_party/prometheus-cpp' 2024-08-06T20:33:29.5780613Z Submodule 'tools/vcpkg' (https://github.com/Microsoft/vcpkg) registered for path 'third_party/opentelemetry-cpp/tools/vcpkg' 2024-08-06T20:33:29.5812268Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/opentelemetry-cpp/third_party/benchmark'... 2024-08-06T20:33:30.0735137Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/opentelemetry-cpp/third_party/googletest'... 2024-08-06T20:33:31.3429679Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/opentelemetry-cpp/third_party/ms-gsl'... 2024-08-06T20:33:31.7202162Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/opentelemetry-cpp/third_party/nlohmann-json'... 2024-08-06T20:33:39.2683563Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/opentelemetry-cpp/third_party/opentelemetry-proto'... 2024-08-06T20:33:39.5323782Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/opentelemetry-cpp/third_party/opentracing-cpp'... 2024-08-06T20:33:39.7509106Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/opentelemetry-cpp/third_party/prometheus-cpp'... 2024-08-06T20:33:40.0846962Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/opentelemetry-cpp/tools/vcpkg'... 2024-08-06T20:33:48.6455543Z Submodule path 'third_party/opentelemetry-cpp/third_party/benchmark': checked out 'd572f4777349d43653b21d6c2fc63020ab326db2' 2024-08-06T20:33:48.6967937Z Submodule path 'third_party/opentelemetry-cpp/third_party/googletest': checked out 'b796f7d44681514f58a683a3a71ff17c94edb0c1' 2024-08-06T20:33:48.7163844Z Submodule path 'third_party/opentelemetry-cpp/third_party/ms-gsl': checked out '6f4529395c5b7c2d661812257cd6780c67e54afa' 2024-08-06T20:33:48.8490299Z Submodule path 'third_party/opentelemetry-cpp/third_party/nlohmann-json': checked out 'bc889afb4c5bf1c0d8ee29ef35eaaf4c8bef8a5d' 2024-08-06T20:33:48.8654077Z Submodule path 'third_party/opentelemetry-cpp/third_party/opentelemetry-proto': checked out '4ca4f0335c63cda7ab31ea7ed70d6553aee14dce' 2024-08-06T20:33:48.8839893Z Submodule path 'third_party/opentelemetry-cpp/third_party/opentracing-cpp': checked out '06b57f48ded1fa3bdd3d4346f6ef29e40e08eaf5' 2024-08-06T20:33:48.9043591Z Submodule path 'third_party/opentelemetry-cpp/third_party/prometheus-cpp': checked out 'c9ffcdda9086ffd9e1283ea7a0276d831f3c8a8d' 2024-08-06T20:33:48.9063201Z Submodule 'civetweb' (https://github.com/civetweb/civetweb.git) registered for path 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/civetweb' 2024-08-06T20:33:48.9066844Z Submodule 'googletest' (https://github.com/google/googletest.git) registered for path 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/googletest' 2024-08-06T20:33:48.9095829Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/civetweb'... 2024-08-06T20:33:51.2116410Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/googletest'... 2024-08-06T20:33:52.8461844Z Submodule path 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/civetweb': checked out 'eefb26f82b233268fc98577d265352720d477ba4' 2024-08-06T20:33:52.9052638Z Submodule path 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/googletest': checked out 'e2239ee6043f73722e7aa812a459f54a28552929' 2024-08-06T20:33:53.6231620Z Submodule path 'third_party/opentelemetry-cpp/tools/vcpkg': checked out '8eb57355a4ffb410a2e94c07b4dca2dffbee8e50' 2024-08-06T20:33:53.6372236Z Submodule path 'third_party/pocketfft': checked out '9d3ab05a7fffbc71a492bc6a17be034e83e8f0fe' 2024-08-06T20:33:53.9930403Z Submodule path 'third_party/protobuf': checked out 'd1eca4e4b421cd2997495c4b4e65cea6be4e9b8a' 2024-08-06T20:33:53.9957165Z Submodule 'third_party/benchmark' (https://github.com/google/benchmark.git) registered for path 'third_party/protobuf/third_party/benchmark' 2024-08-06T20:33:53.9960954Z Submodule 'third_party/googletest' (https://github.com/google/googletest.git) registered for path 'third_party/protobuf/third_party/googletest' 2024-08-06T20:33:53.9992662Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/protobuf/third_party/benchmark'... 2024-08-06T20:33:54.5552315Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/protobuf/third_party/googletest'... 2024-08-06T20:33:55.8463666Z Submodule path 'third_party/protobuf/third_party/benchmark': checked out '5b7683f49e1e9223cf9927b24f6fd3d6bd82e3f8' 2024-08-06T20:33:55.9371295Z Submodule path 'third_party/protobuf/third_party/googletest': checked out '5ec7f0c4a113e2f18ac2c6cc7df51ad6afc24081' 2024-08-06T20:33:55.9478027Z Submodule path 'third_party/psimd': checked out '072586a71b55b7f8c584153d223e95687148a900' 2024-08-06T20:33:55.9635997Z Submodule path 'third_party/pthreadpool': checked out '4fe0e1e183925bf8cfa6aae24237e724a96479b8' 2024-08-06T20:33:56.0115485Z Submodule path 'third_party/pybind11': checked out '941f45bcb51457884fa1afd6e24a67377d70f75c' 2024-08-06T20:33:56.0477101Z Submodule path 'third_party/python-peachpy': checked out 'f45429b087dd7d5bc78bb40dc7cf06425c252d67' 2024-08-06T20:33:56.0998260Z Submodule path 'third_party/sleef': checked out '60e76d2bce17d278b439d9da17177c8f957a9e9b' 2024-08-06T20:33:56.1354951Z Submodule path 'third_party/tensorpipe': checked out '52791a2fd214b2a9dc5759d36725909c1daa7f2e' 2024-08-06T20:33:56.1375510Z Submodule 'third_party/googletest' (https://github.com/google/googletest.git) registered for path 'third_party/tensorpipe/third_party/googletest' 2024-08-06T20:33:56.1379487Z Submodule 'third_party/libnop' (https://github.com/google/libnop.git) registered for path 'third_party/tensorpipe/third_party/libnop' 2024-08-06T20:33:56.1382996Z Submodule 'third_party/libuv' (https://github.com/libuv/libuv.git) registered for path 'third_party/tensorpipe/third_party/libuv' 2024-08-06T20:33:56.1386822Z Submodule 'third_party/pybind11' (https://github.com/pybind/pybind11.git) registered for path 'third_party/tensorpipe/third_party/pybind11' 2024-08-06T20:33:56.1417045Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/tensorpipe/third_party/googletest'... 2024-08-06T20:33:57.4460166Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/tensorpipe/third_party/libnop'... 2024-08-06T20:33:57.7682635Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/tensorpipe/third_party/libuv'... 2024-08-06T20:33:59.2110403Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/tensorpipe/third_party/pybind11'... 2024-08-06T20:34:00.4296090Z Submodule path 'third_party/tensorpipe/third_party/googletest': checked out 'aee0f9d9b5b87796ee8a0ab26b7587ec30e8858e' 2024-08-06T20:34:00.4491569Z Submodule path 'third_party/tensorpipe/third_party/libnop': checked out '910b55815be16109f04f4180e9adee14fb4ce281' 2024-08-06T20:34:00.5373585Z Submodule path 'third_party/tensorpipe/third_party/libuv': checked out '1dff88e5161cba5c59276d2070d2e304e4dcb242' 2024-08-06T20:34:00.5740442Z Submodule path 'third_party/tensorpipe/third_party/pybind11': checked out 'a23996fce38ff6ccfbcdc09f1e63f2c4be5ea2ef' 2024-08-06T20:34:00.5759297Z Submodule 'tools/clang' (https://github.com/wjakob/clang-cindex-python3) registered for path 'third_party/tensorpipe/third_party/pybind11/tools/clang' 2024-08-06T20:34:00.5790926Z Cloning into '/home/ec2-user/actions-runner/_work/pytorch/pytorch/third_party/tensorpipe/third_party/pybind11/tools/clang'... 2024-08-06T20:34:00.8147949Z Submodule path 'third_party/tensorpipe/third_party/pybind11/tools/clang': checked out '6a00cbc4a9b8e68b71caf7f774b3f9c753ae84d5' 2024-08-06T20:34:00.8186608Z [command]/usr/bin/git submodule foreach --recursive git config --local gc.auto 0 2024-08-06T20:34:00.8528159Z Entering 'android/libs/fbjni' 2024-08-06T20:34:00.8576542Z Entering 'third_party/FP16' 2024-08-06T20:34:00.8627025Z Entering 'third_party/FXdiv' 2024-08-06T20:34:00.8676451Z Entering 'third_party/NNPACK' 2024-08-06T20:34:00.8726304Z Entering 'third_party/VulkanMemoryAllocator' 2024-08-06T20:34:00.8775394Z Entering 'third_party/XNNPACK' 2024-08-06T20:34:00.8842928Z Entering 'third_party/benchmark' 2024-08-06T20:34:00.8892067Z Entering 'third_party/cpp-httplib' 2024-08-06T20:34:00.8940990Z Entering 'third_party/cpuinfo' 2024-08-06T20:34:00.8991197Z Entering 'third_party/cudnn_frontend' 2024-08-06T20:34:00.9041888Z Entering 'third_party/cutlass' 2024-08-06T20:34:00.9100250Z Entering 'third_party/eigen' 2024-08-06T20:34:00.9152654Z Entering 'third_party/fbgemm' 2024-08-06T20:34:00.9200900Z Entering 'third_party/fbgemm/third_party/asmjit' 2024-08-06T20:34:00.9248731Z Entering 'third_party/fbgemm/third_party/cpuinfo' 2024-08-06T20:34:00.9295936Z Entering 'third_party/fbgemm/third_party/cutlass' 2024-08-06T20:34:00.9351939Z Entering 'third_party/fbgemm/third_party/googletest' 2024-08-06T20:34:00.9399436Z Entering 'third_party/fbgemm/third_party/hipify_torch' 2024-08-06T20:34:00.9449106Z Entering 'third_party/flatbuffers' 2024-08-06T20:34:00.9500966Z Entering 'third_party/fmt' 2024-08-06T20:34:00.9550872Z Entering 'third_party/foxi' 2024-08-06T20:34:00.9598992Z Entering 'third_party/gemmlowp/gemmlowp' 2024-08-06T20:34:00.9649297Z Entering 'third_party/gloo' 2024-08-06T20:34:00.9697209Z Entering 'third_party/googletest' 2024-08-06T20:34:00.9746824Z Entering 'third_party/ideep' 2024-08-06T20:34:00.9793917Z Entering 'third_party/ideep/mkl-dnn' 2024-08-06T20:34:00.9849744Z Entering 'third_party/ittapi' 2024-08-06T20:34:00.9897736Z Entering 'third_party/kineto' 2024-08-06T20:34:00.9946366Z Entering 'third_party/kineto/libkineto/third_party/dynolog' 2024-08-06T20:34:00.9994440Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/DCGM' 2024-08-06T20:34:01.0045179Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/cpr' 2024-08-06T20:34:01.0093046Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/fmt' 2024-08-06T20:34:01.0142727Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags' 2024-08-06T20:34:01.0189862Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags/doc' 2024-08-06T20:34:01.0240497Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/glog' 2024-08-06T20:34:01.0288046Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/googletest' 2024-08-06T20:34:01.0336485Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/json' 2024-08-06T20:34:01.0385633Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/pfs' 2024-08-06T20:34:01.0437391Z Entering 'third_party/kineto/libkineto/third_party/fmt' 2024-08-06T20:34:01.0484831Z Entering 'third_party/kineto/libkineto/third_party/googletest' 2024-08-06T20:34:01.0535801Z Entering 'third_party/mimalloc' 2024-08-06T20:34:01.0583876Z Entering 'third_party/nccl/nccl' 2024-08-06T20:34:01.0632684Z Entering 'third_party/nlohmann' 2024-08-06T20:34:01.0682350Z Entering 'third_party/onnx' 2024-08-06T20:34:01.0747604Z Entering 'third_party/onnx/third_party/benchmark' 2024-08-06T20:34:01.0794998Z Entering 'third_party/onnx/third_party/pybind11' 2024-08-06T20:34:01.0845748Z Entering 'third_party/opentelemetry-cpp' 2024-08-06T20:34:01.0896182Z Entering 'third_party/opentelemetry-cpp/third_party/benchmark' 2024-08-06T20:34:01.0944499Z Entering 'third_party/opentelemetry-cpp/third_party/googletest' 2024-08-06T20:34:01.0991667Z Entering 'third_party/opentelemetry-cpp/third_party/ms-gsl' 2024-08-06T20:34:01.1039066Z Entering 'third_party/opentelemetry-cpp/third_party/nlohmann-json' 2024-08-06T20:34:01.1087738Z Entering 'third_party/opentelemetry-cpp/third_party/opentelemetry-proto' 2024-08-06T20:34:01.1136063Z Entering 'third_party/opentelemetry-cpp/third_party/opentracing-cpp' 2024-08-06T20:34:01.1183179Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp' 2024-08-06T20:34:01.1230901Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/civetweb' 2024-08-06T20:34:01.1279881Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/googletest' 2024-08-06T20:34:01.1330195Z Entering 'third_party/opentelemetry-cpp/tools/vcpkg' 2024-08-06T20:34:01.1402035Z Entering 'third_party/pocketfft' 2024-08-06T20:34:01.1450092Z Entering 'third_party/protobuf' 2024-08-06T20:34:01.1501792Z Entering 'third_party/protobuf/third_party/benchmark' 2024-08-06T20:34:01.1549771Z Entering 'third_party/protobuf/third_party/googletest' 2024-08-06T20:34:01.1598887Z Entering 'third_party/psimd' 2024-08-06T20:34:01.1647712Z Entering 'third_party/pthreadpool' 2024-08-06T20:34:01.1695339Z Entering 'third_party/pybind11' 2024-08-06T20:34:01.1743702Z Entering 'third_party/python-peachpy' 2024-08-06T20:34:01.1791500Z Entering 'third_party/sleef' 2024-08-06T20:34:01.1841229Z Entering 'third_party/tensorpipe' 2024-08-06T20:34:01.1889752Z Entering 'third_party/tensorpipe/third_party/googletest' 2024-08-06T20:34:01.1938049Z Entering 'third_party/tensorpipe/third_party/libnop' 2024-08-06T20:34:01.1985262Z Entering 'third_party/tensorpipe/third_party/libuv' 2024-08-06T20:34:01.2034903Z Entering 'third_party/tensorpipe/third_party/pybind11' 2024-08-06T20:34:01.2081229Z Entering 'third_party/tensorpipe/third_party/pybind11/tools/clang' 2024-08-06T20:34:01.2145336Z ##[endgroup] 2024-08-06T20:34:01.2147977Z ##[group]Persisting credentials for submodules 2024-08-06T20:34:01.2152952Z [command]/usr/bin/git submodule foreach --recursive sh -c "git config --local --name-only --get-regexp 'url\.https\:\/\/github\.com\/\.insteadOf' && git config --local --unset-all 'url.https://github.com/.insteadOf' || :" 2024-08-06T20:34:01.2491614Z Entering 'android/libs/fbjni' 2024-08-06T20:34:01.2555642Z Entering 'third_party/FP16' 2024-08-06T20:34:01.2618263Z Entering 'third_party/FXdiv' 2024-08-06T20:34:01.2679535Z Entering 'third_party/NNPACK' 2024-08-06T20:34:01.2741795Z Entering 'third_party/VulkanMemoryAllocator' 2024-08-06T20:34:01.2804984Z Entering 'third_party/XNNPACK' 2024-08-06T20:34:01.2885676Z Entering 'third_party/benchmark' 2024-08-06T20:34:01.2948561Z Entering 'third_party/cpp-httplib' 2024-08-06T20:34:01.3012440Z Entering 'third_party/cpuinfo' 2024-08-06T20:34:01.3074759Z Entering 'third_party/cudnn_frontend' 2024-08-06T20:34:01.3137647Z Entering 'third_party/cutlass' 2024-08-06T20:34:01.3209915Z Entering 'third_party/eigen' 2024-08-06T20:34:01.3273887Z Entering 'third_party/fbgemm' 2024-08-06T20:34:01.3336258Z Entering 'third_party/fbgemm/third_party/asmjit' 2024-08-06T20:34:01.3396864Z Entering 'third_party/fbgemm/third_party/cpuinfo' 2024-08-06T20:34:01.3459446Z Entering 'third_party/fbgemm/third_party/cutlass' 2024-08-06T20:34:01.3528975Z Entering 'third_party/fbgemm/third_party/googletest' 2024-08-06T20:34:01.3590327Z Entering 'third_party/fbgemm/third_party/hipify_torch' 2024-08-06T20:34:01.3655492Z Entering 'third_party/flatbuffers' 2024-08-06T20:34:01.3721443Z Entering 'third_party/fmt' 2024-08-06T20:34:01.3782556Z Entering 'third_party/foxi' 2024-08-06T20:34:01.3844827Z Entering 'third_party/gemmlowp/gemmlowp' 2024-08-06T20:34:01.3908771Z Entering 'third_party/gloo' 2024-08-06T20:34:01.3970272Z Entering 'third_party/googletest' 2024-08-06T20:34:01.4034705Z Entering 'third_party/ideep' 2024-08-06T20:34:01.4095036Z Entering 'third_party/ideep/mkl-dnn' 2024-08-06T20:34:01.4165625Z Entering 'third_party/ittapi' 2024-08-06T20:34:01.4227956Z Entering 'third_party/kineto' 2024-08-06T20:34:01.4289725Z Entering 'third_party/kineto/libkineto/third_party/dynolog' 2024-08-06T20:34:01.4351668Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/DCGM' 2024-08-06T20:34:01.4416878Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/cpr' 2024-08-06T20:34:01.4478608Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/fmt' 2024-08-06T20:34:01.4542224Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags' 2024-08-06T20:34:01.4603048Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags/doc' 2024-08-06T20:34:01.4668317Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/glog' 2024-08-06T20:34:01.4731827Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/googletest' 2024-08-06T20:34:01.4794045Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/json' 2024-08-06T20:34:01.4858214Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/pfs' 2024-08-06T20:34:01.4922442Z Entering 'third_party/kineto/libkineto/third_party/fmt' 2024-08-06T20:34:01.4983943Z Entering 'third_party/kineto/libkineto/third_party/googletest' 2024-08-06T20:34:01.5048113Z Entering 'third_party/mimalloc' 2024-08-06T20:34:01.5111458Z Entering 'third_party/nccl/nccl' 2024-08-06T20:34:01.5173873Z Entering 'third_party/nlohmann' 2024-08-06T20:34:01.5239166Z Entering 'third_party/onnx' 2024-08-06T20:34:01.5318785Z Entering 'third_party/onnx/third_party/benchmark' 2024-08-06T20:34:01.5380235Z Entering 'third_party/onnx/third_party/pybind11' 2024-08-06T20:34:01.5445023Z Entering 'third_party/opentelemetry-cpp' 2024-08-06T20:34:01.5509545Z Entering 'third_party/opentelemetry-cpp/third_party/benchmark' 2024-08-06T20:34:01.5571009Z Entering 'third_party/opentelemetry-cpp/third_party/googletest' 2024-08-06T20:34:01.5633668Z Entering 'third_party/opentelemetry-cpp/third_party/ms-gsl' 2024-08-06T20:34:01.5694523Z Entering 'third_party/opentelemetry-cpp/third_party/nlohmann-json' 2024-08-06T20:34:01.5758843Z Entering 'third_party/opentelemetry-cpp/third_party/opentelemetry-proto' 2024-08-06T20:34:01.5820184Z Entering 'third_party/opentelemetry-cpp/third_party/opentracing-cpp' 2024-08-06T20:34:01.5881744Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp' 2024-08-06T20:34:01.5942952Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/civetweb' 2024-08-06T20:34:01.6007814Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/googletest' 2024-08-06T20:34:01.6071733Z Entering 'third_party/opentelemetry-cpp/tools/vcpkg' 2024-08-06T20:34:01.6157700Z Entering 'third_party/pocketfft' 2024-08-06T20:34:01.6220394Z Entering 'third_party/protobuf' 2024-08-06T20:34:01.6285850Z Entering 'third_party/protobuf/third_party/benchmark' 2024-08-06T20:34:01.6348595Z Entering 'third_party/protobuf/third_party/googletest' 2024-08-06T20:34:01.6414039Z Entering 'third_party/psimd' 2024-08-06T20:34:01.6475900Z Entering 'third_party/pthreadpool' 2024-08-06T20:34:01.6538517Z Entering 'third_party/pybind11' 2024-08-06T20:34:01.6601129Z Entering 'third_party/python-peachpy' 2024-08-06T20:34:01.6664324Z Entering 'third_party/sleef' 2024-08-06T20:34:01.6728046Z Entering 'third_party/tensorpipe' 2024-08-06T20:34:01.6789857Z Entering 'third_party/tensorpipe/third_party/googletest' 2024-08-06T20:34:01.6852624Z Entering 'third_party/tensorpipe/third_party/libnop' 2024-08-06T20:34:01.6915746Z Entering 'third_party/tensorpipe/third_party/libuv' 2024-08-06T20:34:01.6977601Z Entering 'third_party/tensorpipe/third_party/pybind11' 2024-08-06T20:34:01.7039043Z Entering 'third_party/tensorpipe/third_party/pybind11/tools/clang' 2024-08-06T20:34:01.7131856Z [command]/usr/bin/git submodule foreach --recursive sh -c "git config --local 'http.https://github.com/.extraheader' 'AUTHORIZATION: basic ***' && git config --local --show-origin --name-only --get-regexp remote.origin.url" 2024-08-06T20:34:01.7477370Z Entering 'android/libs/fbjni' 2024-08-06T20:34:01.7536405Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/android/libs/fbjni/config remote.origin.url 2024-08-06T20:34:01.7555749Z Entering 'third_party/FP16' 2024-08-06T20:34:01.7617640Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/NNPACK_deps/FP16/config remote.origin.url 2024-08-06T20:34:01.7636464Z Entering 'third_party/FXdiv' 2024-08-06T20:34:01.7696414Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/NNPACK_deps/FXdiv/config remote.origin.url 2024-08-06T20:34:01.7715584Z Entering 'third_party/NNPACK' 2024-08-06T20:34:01.7774070Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/NNPACK/config remote.origin.url 2024-08-06T20:34:01.7793142Z Entering 'third_party/VulkanMemoryAllocator' 2024-08-06T20:34:01.7853220Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/VulkanMemoryAllocator/config remote.origin.url 2024-08-06T20:34:01.7872012Z Entering 'third_party/XNNPACK' 2024-08-06T20:34:01.7932021Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/XNNPACK/config remote.origin.url 2024-08-06T20:34:01.7969455Z Entering 'third_party/benchmark' 2024-08-06T20:34:01.8029849Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/benchmark/config remote.origin.url 2024-08-06T20:34:01.8048728Z Entering 'third_party/cpp-httplib' 2024-08-06T20:34:01.8110089Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/cpp-httplib/config remote.origin.url 2024-08-06T20:34:01.8128817Z Entering 'third_party/cpuinfo' 2024-08-06T20:34:01.8187566Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/cpuinfo/config remote.origin.url 2024-08-06T20:34:01.8208300Z Entering 'third_party/cudnn_frontend' 2024-08-06T20:34:01.8266072Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/cudnn_frontend/config remote.origin.url 2024-08-06T20:34:01.8285911Z Entering 'third_party/cutlass' 2024-08-06T20:34:01.8346156Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/cutlass/config remote.origin.url 2024-08-06T20:34:01.8373149Z Entering 'third_party/eigen' 2024-08-06T20:34:01.8433698Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/eigen/config remote.origin.url 2024-08-06T20:34:01.8455206Z Entering 'third_party/fbgemm' 2024-08-06T20:34:01.8515866Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/fbgemm/config remote.origin.url 2024-08-06T20:34:01.8534366Z Entering 'third_party/fbgemm/third_party/asmjit' 2024-08-06T20:34:01.8593332Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/fbgemm/modules/third_party/asmjit/config remote.origin.url 2024-08-06T20:34:01.8613077Z Entering 'third_party/fbgemm/third_party/cpuinfo' 2024-08-06T20:34:01.8671923Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/fbgemm/modules/third_party/cpuinfo/config remote.origin.url 2024-08-06T20:34:01.8690496Z Entering 'third_party/fbgemm/third_party/cutlass' 2024-08-06T20:34:01.8750142Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/fbgemm/modules/third_party/cutlass/config remote.origin.url 2024-08-06T20:34:01.8776391Z Entering 'third_party/fbgemm/third_party/googletest' 2024-08-06T20:34:01.8835885Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/fbgemm/modules/third_party/googletest/config remote.origin.url 2024-08-06T20:34:01.8854335Z Entering 'third_party/fbgemm/third_party/hipify_torch' 2024-08-06T20:34:01.8914269Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/fbgemm/modules/third_party/hipify_torch/config remote.origin.url 2024-08-06T20:34:01.8934176Z Entering 'third_party/flatbuffers' 2024-08-06T20:34:01.8993799Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/flatbuffers/config remote.origin.url 2024-08-06T20:34:01.9017187Z Entering 'third_party/fmt' 2024-08-06T20:34:01.9075409Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/fmt/config remote.origin.url 2024-08-06T20:34:01.9094240Z Entering 'third_party/foxi' 2024-08-06T20:34:01.9154752Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/foxi/config remote.origin.url 2024-08-06T20:34:01.9173416Z Entering 'third_party/gemmlowp/gemmlowp' 2024-08-06T20:34:01.9232272Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/gemmlowp/gemmlowp/config remote.origin.url 2024-08-06T20:34:01.9251558Z Entering 'third_party/gloo' 2024-08-06T20:34:01.9310706Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/gloo/config remote.origin.url 2024-08-06T20:34:01.9329715Z Entering 'third_party/googletest' 2024-08-06T20:34:01.9389042Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/googletest/config remote.origin.url 2024-08-06T20:34:01.9409105Z Entering 'third_party/ideep' 2024-08-06T20:34:01.9467850Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/ideep/config remote.origin.url 2024-08-06T20:34:01.9485437Z Entering 'third_party/ideep/mkl-dnn' 2024-08-06T20:34:01.9544328Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/ideep/modules/mkl-dnn/config remote.origin.url 2024-08-06T20:34:01.9571353Z Entering 'third_party/ittapi' 2024-08-06T20:34:01.9630556Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/ittapi/config remote.origin.url 2024-08-06T20:34:01.9649643Z Entering 'third_party/kineto' 2024-08-06T20:34:01.9711102Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/kineto/config remote.origin.url 2024-08-06T20:34:01.9729345Z Entering 'third_party/kineto/libkineto/third_party/dynolog' 2024-08-06T20:34:01.9789330Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/kineto/modules/libkineto/third_party/dynolog/config remote.origin.url 2024-08-06T20:34:01.9809470Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/DCGM' 2024-08-06T20:34:01.9870174Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/kineto/modules/libkineto/third_party/dynolog/modules/third_party/DCGM/config remote.origin.url 2024-08-06T20:34:01.9890151Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/cpr' 2024-08-06T20:34:01.9953883Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/kineto/modules/libkineto/third_party/dynolog/modules/third_party/cpr/config remote.origin.url 2024-08-06T20:34:01.9972866Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/fmt' 2024-08-06T20:34:02.0034861Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/kineto/modules/libkineto/third_party/dynolog/modules/third_party/fmt/config remote.origin.url 2024-08-06T20:34:02.0053999Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags' 2024-08-06T20:34:02.0116651Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/kineto/modules/libkineto/third_party/dynolog/modules/third_party/gflags/config remote.origin.url 2024-08-06T20:34:02.0134257Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags/doc' 2024-08-06T20:34:02.0193379Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/kineto/modules/libkineto/third_party/dynolog/modules/third_party/gflags/modules/doc/config remote.origin.url 2024-08-06T20:34:02.0214756Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/glog' 2024-08-06T20:34:02.0273981Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/kineto/modules/libkineto/third_party/dynolog/modules/third_party/glog/config remote.origin.url 2024-08-06T20:34:02.0292813Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/googletest' 2024-08-06T20:34:02.0353252Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/kineto/modules/libkineto/third_party/dynolog/modules/third_party/googletest/config remote.origin.url 2024-08-06T20:34:02.0372544Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/json' 2024-08-06T20:34:02.0432413Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/kineto/modules/libkineto/third_party/dynolog/modules/third_party/json/config remote.origin.url 2024-08-06T20:34:02.0452539Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/pfs' 2024-08-06T20:34:02.0512203Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/kineto/modules/libkineto/third_party/dynolog/modules/third_party/pfs/config remote.origin.url 2024-08-06T20:34:02.0532623Z Entering 'third_party/kineto/libkineto/third_party/fmt' 2024-08-06T20:34:02.0591562Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/kineto/modules/libkineto/third_party/fmt/config remote.origin.url 2024-08-06T20:34:02.0610573Z Entering 'third_party/kineto/libkineto/third_party/googletest' 2024-08-06T20:34:02.0668769Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/kineto/modules/libkineto/third_party/googletest/config remote.origin.url 2024-08-06T20:34:02.0689318Z Entering 'third_party/mimalloc' 2024-08-06T20:34:02.0749946Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/mimalloc/config remote.origin.url 2024-08-06T20:34:02.0768958Z Entering 'third_party/nccl/nccl' 2024-08-06T20:34:02.0827602Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/nccl/nccl/config remote.origin.url 2024-08-06T20:34:02.0846722Z Entering 'third_party/nlohmann' 2024-08-06T20:34:02.0905493Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/nlohmann/config remote.origin.url 2024-08-06T20:34:02.0925885Z Entering 'third_party/onnx' 2024-08-06T20:34:02.0984526Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/onnx/config remote.origin.url 2024-08-06T20:34:02.1021838Z Entering 'third_party/onnx/third_party/benchmark' 2024-08-06T20:34:02.1080199Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/onnx/modules/third_party/benchmark/config remote.origin.url 2024-08-06T20:34:02.1098634Z Entering 'third_party/onnx/third_party/pybind11' 2024-08-06T20:34:02.1160196Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/onnx/modules/third_party/pybind11/config remote.origin.url 2024-08-06T20:34:02.1181355Z Entering 'third_party/opentelemetry-cpp' 2024-08-06T20:34:02.1242196Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/opentelemetry-cpp/config remote.origin.url 2024-08-06T20:34:02.1262279Z Entering 'third_party/opentelemetry-cpp/third_party/benchmark' 2024-08-06T20:34:02.1320890Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/opentelemetry-cpp/modules/third_party/benchmark/config remote.origin.url 2024-08-06T20:34:02.1339124Z Entering 'third_party/opentelemetry-cpp/third_party/googletest' 2024-08-06T20:34:02.1398617Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/opentelemetry-cpp/modules/third_party/googletest/config remote.origin.url 2024-08-06T20:34:02.1418493Z Entering 'third_party/opentelemetry-cpp/third_party/ms-gsl' 2024-08-06T20:34:02.1476236Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/opentelemetry-cpp/modules/third_party/ms-gsl/config remote.origin.url 2024-08-06T20:34:02.1494382Z Entering 'third_party/opentelemetry-cpp/third_party/nlohmann-json' 2024-08-06T20:34:02.1554358Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/opentelemetry-cpp/modules/third_party/nlohmann-json/config remote.origin.url 2024-08-06T20:34:02.1574929Z Entering 'third_party/opentelemetry-cpp/third_party/opentelemetry-proto' 2024-08-06T20:34:02.1634585Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/opentelemetry-cpp/modules/third_party/opentelemetry-proto/config remote.origin.url 2024-08-06T20:34:02.1652545Z Entering 'third_party/opentelemetry-cpp/third_party/opentracing-cpp' 2024-08-06T20:34:02.1712071Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/opentelemetry-cpp/modules/third_party/opentracing-cpp/config remote.origin.url 2024-08-06T20:34:02.1730116Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp' 2024-08-06T20:34:02.1789671Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/opentelemetry-cpp/modules/third_party/prometheus-cpp/config remote.origin.url 2024-08-06T20:34:02.1808027Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/civetweb' 2024-08-06T20:34:02.1866525Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/opentelemetry-cpp/modules/third_party/prometheus-cpp/modules/civetweb/config remote.origin.url 2024-08-06T20:34:02.1887116Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/googletest' 2024-08-06T20:34:02.1946129Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/opentelemetry-cpp/modules/third_party/prometheus-cpp/modules/googletest/config remote.origin.url 2024-08-06T20:34:02.1966655Z Entering 'third_party/opentelemetry-cpp/tools/vcpkg' 2024-08-06T20:34:02.2025122Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/opentelemetry-cpp/modules/tools/vcpkg/config remote.origin.url 2024-08-06T20:34:02.2067154Z Entering 'third_party/pocketfft' 2024-08-06T20:34:02.2127241Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/pocketfft/config remote.origin.url 2024-08-06T20:34:02.2145965Z Entering 'third_party/protobuf' 2024-08-06T20:34:02.2207374Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/protobuf/config remote.origin.url 2024-08-06T20:34:02.2228222Z Entering 'third_party/protobuf/third_party/benchmark' 2024-08-06T20:34:02.2286823Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/protobuf/modules/third_party/benchmark/config remote.origin.url 2024-08-06T20:34:02.2305650Z Entering 'third_party/protobuf/third_party/googletest' 2024-08-06T20:34:02.2365310Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/protobuf/modules/third_party/googletest/config remote.origin.url 2024-08-06T20:34:02.2385984Z Entering 'third_party/psimd' 2024-08-06T20:34:02.2446623Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/NNPACK_deps/psimd/config remote.origin.url 2024-08-06T20:34:02.2465364Z Entering 'third_party/pthreadpool' 2024-08-06T20:34:02.2524126Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/NNPACK_deps/pthreadpool/config remote.origin.url 2024-08-06T20:34:02.2542648Z Entering 'third_party/pybind11' 2024-08-06T20:34:02.2602066Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/pybind11/config remote.origin.url 2024-08-06T20:34:02.2622437Z Entering 'third_party/python-peachpy' 2024-08-06T20:34:02.2681062Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/python-peachpy/config remote.origin.url 2024-08-06T20:34:02.2699596Z Entering 'third_party/sleef' 2024-08-06T20:34:02.2758997Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/sleef/config remote.origin.url 2024-08-06T20:34:02.2777691Z Entering 'third_party/tensorpipe' 2024-08-06T20:34:02.2838244Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/tensorpipe/config remote.origin.url 2024-08-06T20:34:02.2856243Z Entering 'third_party/tensorpipe/third_party/googletest' 2024-08-06T20:34:02.2915235Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/tensorpipe/modules/third_party/googletest/config remote.origin.url 2024-08-06T20:34:02.2933445Z Entering 'third_party/tensorpipe/third_party/libnop' 2024-08-06T20:34:02.2991196Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/tensorpipe/modules/third_party/libnop/config remote.origin.url 2024-08-06T20:34:02.3011198Z Entering 'third_party/tensorpipe/third_party/libuv' 2024-08-06T20:34:02.3069015Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/tensorpipe/modules/third_party/libuv/config remote.origin.url 2024-08-06T20:34:02.3087732Z Entering 'third_party/tensorpipe/third_party/pybind11' 2024-08-06T20:34:02.3147714Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/tensorpipe/modules/third_party/pybind11/config remote.origin.url 2024-08-06T20:34:02.3164783Z Entering 'third_party/tensorpipe/third_party/pybind11/tools/clang' 2024-08-06T20:34:02.3224754Z file:/home/ec2-user/actions-runner/_work/pytorch/pytorch/.git/modules/third_party/tensorpipe/modules/third_party/pybind11/modules/tools/clang/config remote.origin.url 2024-08-06T20:34:02.4547675Z [command]/usr/bin/git submodule foreach --recursive git config --local --add 'url.https://github.com/.insteadOf' 'git@github.com:' 2024-08-06T20:34:02.4901118Z Entering 'android/libs/fbjni' 2024-08-06T20:34:02.4950127Z Entering 'third_party/FP16' 2024-08-06T20:34:02.4998829Z Entering 'third_party/FXdiv' 2024-08-06T20:34:02.5048995Z Entering 'third_party/NNPACK' 2024-08-06T20:34:02.5097910Z Entering 'third_party/VulkanMemoryAllocator' 2024-08-06T20:34:02.5148588Z Entering 'third_party/XNNPACK' 2024-08-06T20:34:02.5216139Z Entering 'third_party/benchmark' 2024-08-06T20:34:02.5265193Z Entering 'third_party/cpp-httplib' 2024-08-06T20:34:02.5315558Z Entering 'third_party/cpuinfo' 2024-08-06T20:34:02.5364981Z Entering 'third_party/cudnn_frontend' 2024-08-06T20:34:02.5415126Z Entering 'third_party/cutlass' 2024-08-06T20:34:02.5471849Z Entering 'third_party/eigen' 2024-08-06T20:34:02.5523472Z Entering 'third_party/fbgemm' 2024-08-06T20:34:02.5572934Z Entering 'third_party/fbgemm/third_party/asmjit' 2024-08-06T20:34:02.5620599Z Entering 'third_party/fbgemm/third_party/cpuinfo' 2024-08-06T20:34:02.5668510Z Entering 'third_party/fbgemm/third_party/cutlass' 2024-08-06T20:34:02.5723412Z Entering 'third_party/fbgemm/third_party/googletest' 2024-08-06T20:34:02.5770670Z Entering 'third_party/fbgemm/third_party/hipify_torch' 2024-08-06T20:34:02.5820006Z Entering 'third_party/flatbuffers' 2024-08-06T20:34:02.5871699Z Entering 'third_party/fmt' 2024-08-06T20:34:02.5920162Z Entering 'third_party/foxi' 2024-08-06T20:34:02.5969708Z Entering 'third_party/gemmlowp/gemmlowp' 2024-08-06T20:34:02.6018510Z Entering 'third_party/gloo' 2024-08-06T20:34:02.6068536Z Entering 'third_party/googletest' 2024-08-06T20:34:02.6117978Z Entering 'third_party/ideep' 2024-08-06T20:34:02.6165969Z Entering 'third_party/ideep/mkl-dnn' 2024-08-06T20:34:02.6220932Z Entering 'third_party/ittapi' 2024-08-06T20:34:02.6269490Z Entering 'third_party/kineto' 2024-08-06T20:34:02.6319448Z Entering 'third_party/kineto/libkineto/third_party/dynolog' 2024-08-06T20:34:02.6368307Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/DCGM' 2024-08-06T20:34:02.6418714Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/cpr' 2024-08-06T20:34:02.6467921Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/fmt' 2024-08-06T20:34:02.6518196Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags' 2024-08-06T20:34:02.6565715Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags/doc' 2024-08-06T20:34:02.6617884Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/glog' 2024-08-06T20:34:02.6666031Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/googletest' 2024-08-06T20:34:02.6716846Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/json' 2024-08-06T20:34:02.6766270Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/pfs' 2024-08-06T20:34:02.6817124Z Entering 'third_party/kineto/libkineto/third_party/fmt' 2024-08-06T20:34:02.6864572Z Entering 'third_party/kineto/libkineto/third_party/googletest' 2024-08-06T20:34:02.6914582Z Entering 'third_party/mimalloc' 2024-08-06T20:34:02.6964382Z Entering 'third_party/nccl/nccl' 2024-08-06T20:34:02.7014626Z Entering 'third_party/nlohmann' 2024-08-06T20:34:02.7064592Z Entering 'third_party/onnx' 2024-08-06T20:34:02.7131902Z Entering 'third_party/onnx/third_party/benchmark' 2024-08-06T20:34:02.7179785Z Entering 'third_party/onnx/third_party/pybind11' 2024-08-06T20:34:02.7230932Z Entering 'third_party/opentelemetry-cpp' 2024-08-06T20:34:02.7281399Z Entering 'third_party/opentelemetry-cpp/third_party/benchmark' 2024-08-06T20:34:02.7329288Z Entering 'third_party/opentelemetry-cpp/third_party/googletest' 2024-08-06T20:34:02.7376161Z Entering 'third_party/opentelemetry-cpp/third_party/ms-gsl' 2024-08-06T20:34:02.7425235Z Entering 'third_party/opentelemetry-cpp/third_party/nlohmann-json' 2024-08-06T20:34:02.7473572Z Entering 'third_party/opentelemetry-cpp/third_party/opentelemetry-proto' 2024-08-06T20:34:02.7522262Z Entering 'third_party/opentelemetry-cpp/third_party/opentracing-cpp' 2024-08-06T20:34:02.7569033Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp' 2024-08-06T20:34:02.7616695Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/civetweb' 2024-08-06T20:34:02.7666418Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/googletest' 2024-08-06T20:34:02.7716964Z Entering 'third_party/opentelemetry-cpp/tools/vcpkg' 2024-08-06T20:34:02.7787332Z Entering 'third_party/pocketfft' 2024-08-06T20:34:02.7837263Z Entering 'third_party/protobuf' 2024-08-06T20:34:02.7891886Z Entering 'third_party/protobuf/third_party/benchmark' 2024-08-06T20:34:02.7942324Z Entering 'third_party/protobuf/third_party/googletest' 2024-08-06T20:34:02.7992391Z Entering 'third_party/psimd' 2024-08-06T20:34:02.8042915Z Entering 'third_party/pthreadpool' 2024-08-06T20:34:02.8091297Z Entering 'third_party/pybind11' 2024-08-06T20:34:02.8141847Z Entering 'third_party/python-peachpy' 2024-08-06T20:34:02.8189793Z Entering 'third_party/sleef' 2024-08-06T20:34:02.8240234Z Entering 'third_party/tensorpipe' 2024-08-06T20:34:02.8288102Z Entering 'third_party/tensorpipe/third_party/googletest' 2024-08-06T20:34:02.8336321Z Entering 'third_party/tensorpipe/third_party/libnop' 2024-08-06T20:34:02.8384125Z Entering 'third_party/tensorpipe/third_party/libuv' 2024-08-06T20:34:02.8432409Z Entering 'third_party/tensorpipe/third_party/pybind11' 2024-08-06T20:34:02.8478731Z Entering 'third_party/tensorpipe/third_party/pybind11/tools/clang' 2024-08-06T20:34:02.8547475Z [command]/usr/bin/git submodule foreach --recursive git config --local --add 'url.https://github.com/.insteadOf' 'org-21003710@github.com:' 2024-08-06T20:34:02.8892534Z Entering 'android/libs/fbjni' 2024-08-06T20:34:02.8942517Z Entering 'third_party/FP16' 2024-08-06T20:34:02.8991660Z Entering 'third_party/FXdiv' 2024-08-06T20:34:02.9040635Z Entering 'third_party/NNPACK' 2024-08-06T20:34:02.9089829Z Entering 'third_party/VulkanMemoryAllocator' 2024-08-06T20:34:02.9138909Z Entering 'third_party/XNNPACK' 2024-08-06T20:34:02.9206645Z Entering 'third_party/benchmark' 2024-08-06T20:34:02.9255220Z Entering 'third_party/cpp-httplib' 2024-08-06T20:34:02.9303381Z Entering 'third_party/cpuinfo' 2024-08-06T20:34:02.9353292Z Entering 'third_party/cudnn_frontend' 2024-08-06T20:34:02.9401368Z Entering 'third_party/cutlass' 2024-08-06T20:34:02.9460651Z Entering 'third_party/eigen' 2024-08-06T20:34:02.9512524Z Entering 'third_party/fbgemm' 2024-08-06T20:34:02.9561382Z Entering 'third_party/fbgemm/third_party/asmjit' 2024-08-06T20:34:02.9608938Z Entering 'third_party/fbgemm/third_party/cpuinfo' 2024-08-06T20:34:02.9656416Z Entering 'third_party/fbgemm/third_party/cutlass' 2024-08-06T20:34:02.9711853Z Entering 'third_party/fbgemm/third_party/googletest' 2024-08-06T20:34:02.9759655Z Entering 'third_party/fbgemm/third_party/hipify_torch' 2024-08-06T20:34:02.9809769Z Entering 'third_party/flatbuffers' 2024-08-06T20:34:02.9861374Z Entering 'third_party/fmt' 2024-08-06T20:34:02.9910687Z Entering 'third_party/foxi' 2024-08-06T20:34:02.9958701Z Entering 'third_party/gemmlowp/gemmlowp' 2024-08-06T20:34:03.0007530Z Entering 'third_party/gloo' 2024-08-06T20:34:03.0056406Z Entering 'third_party/googletest' 2024-08-06T20:34:03.0105554Z Entering 'third_party/ideep' 2024-08-06T20:34:03.0153186Z Entering 'third_party/ideep/mkl-dnn' 2024-08-06T20:34:03.0210312Z Entering 'third_party/ittapi' 2024-08-06T20:34:03.0259436Z Entering 'third_party/kineto' 2024-08-06T20:34:03.0308236Z Entering 'third_party/kineto/libkineto/third_party/dynolog' 2024-08-06T20:34:03.0356014Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/DCGM' 2024-08-06T20:34:03.0406264Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/cpr' 2024-08-06T20:34:03.0454043Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/fmt' 2024-08-06T20:34:03.0502623Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags' 2024-08-06T20:34:03.0551522Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags/doc' 2024-08-06T20:34:03.0601813Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/glog' 2024-08-06T20:34:03.0651628Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/googletest' 2024-08-06T20:34:03.0699580Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/json' 2024-08-06T20:34:03.0750272Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/pfs' 2024-08-06T20:34:03.0800480Z Entering 'third_party/kineto/libkineto/third_party/fmt' 2024-08-06T20:34:03.0848993Z Entering 'third_party/kineto/libkineto/third_party/googletest' 2024-08-06T20:34:03.0898051Z Entering 'third_party/mimalloc' 2024-08-06T20:34:03.0947909Z Entering 'third_party/nccl/nccl' 2024-08-06T20:34:03.0996420Z Entering 'third_party/nlohmann' 2024-08-06T20:34:03.1048457Z Entering 'third_party/onnx' 2024-08-06T20:34:03.1114460Z Entering 'third_party/onnx/third_party/benchmark' 2024-08-06T20:34:03.1161941Z Entering 'third_party/onnx/third_party/pybind11' 2024-08-06T20:34:03.1213826Z Entering 'third_party/opentelemetry-cpp' 2024-08-06T20:34:03.1264348Z Entering 'third_party/opentelemetry-cpp/third_party/benchmark' 2024-08-06T20:34:03.1312315Z Entering 'third_party/opentelemetry-cpp/third_party/googletest' 2024-08-06T20:34:03.1359351Z Entering 'third_party/opentelemetry-cpp/third_party/ms-gsl' 2024-08-06T20:34:03.1406358Z Entering 'third_party/opentelemetry-cpp/third_party/nlohmann-json' 2024-08-06T20:34:03.1455456Z Entering 'third_party/opentelemetry-cpp/third_party/opentelemetry-proto' 2024-08-06T20:34:03.1502212Z Entering 'third_party/opentelemetry-cpp/third_party/opentracing-cpp' 2024-08-06T20:34:03.1550877Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp' 2024-08-06T20:34:03.1597076Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/civetweb' 2024-08-06T20:34:03.1648161Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/googletest' 2024-08-06T20:34:03.1698000Z Entering 'third_party/opentelemetry-cpp/tools/vcpkg' 2024-08-06T20:34:03.1767646Z Entering 'third_party/pocketfft' 2024-08-06T20:34:03.1816544Z Entering 'third_party/protobuf' 2024-08-06T20:34:03.1867778Z Entering 'third_party/protobuf/third_party/benchmark' 2024-08-06T20:34:03.1917241Z Entering 'third_party/protobuf/third_party/googletest' 2024-08-06T20:34:03.1966629Z Entering 'third_party/psimd' 2024-08-06T20:34:03.2015487Z Entering 'third_party/pthreadpool' 2024-08-06T20:34:03.2063320Z Entering 'third_party/pybind11' 2024-08-06T20:34:03.2112160Z Entering 'third_party/python-peachpy' 2024-08-06T20:34:03.2160282Z Entering 'third_party/sleef' 2024-08-06T20:34:03.2209329Z Entering 'third_party/tensorpipe' 2024-08-06T20:34:03.2257074Z Entering 'third_party/tensorpipe/third_party/googletest' 2024-08-06T20:34:03.2303407Z Entering 'third_party/tensorpipe/third_party/libnop' 2024-08-06T20:34:03.2350954Z Entering 'third_party/tensorpipe/third_party/libuv' 2024-08-06T20:34:03.2397538Z Entering 'third_party/tensorpipe/third_party/pybind11' 2024-08-06T20:34:03.2444838Z Entering 'third_party/tensorpipe/third_party/pybind11/tools/clang' 2024-08-06T20:34:03.2507366Z ##[endgroup] 2024-08-06T20:34:03.2564234Z [command]/usr/bin/git log -1 --format='%H' 2024-08-06T20:34:03.2597412Z 'b9d86fa89636e301796d4201f36d86c73f6e49bc' 2024-08-06T20:34:03.2789265Z Prepare all required actions 2024-08-06T20:34:03.2789954Z Getting action download info 2024-08-06T20:34:03.4195551Z ##[group]Run ./.github/actions/setup-linux 2024-08-06T20:34:03.4196166Z env: 2024-08-06T20:34:03.4196442Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:34:03.4196791Z ##[endgroup] 2024-08-06T20:34:03.4275653Z ##[group]Run set -euo pipefail 2024-08-06T20:34:03.4276091Z set -euo pipefail 2024-08-06T20:34:03.4276464Z function get_ec2_metadata() { 2024-08-06T20:34:03.4276950Z  # Pulled from instance metadata endpoint for EC2 2024-08-06T20:34:03.4277734Z  # see https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/instancedata-data-retrieval.html 2024-08-06T20:34:03.4278423Z  category=$1 2024-08-06T20:34:03.4278888Z  # If it is GCP runner (runner name contains gcp), do not run this 2024-08-06T20:34:03.4279449Z  runner_name_str=i-0de4eefb8fab684da 2024-08-06T20:34:03.4279937Z  if [[ -f /.inarc ]]; then 2024-08-06T20:34:03.4280393Z  echo "ARC Runner, no info on ec2 metadata" 2024-08-06T20:34:03.4280879Z  elif [[ $runner_name_str == *"gcp"* ]]; then 2024-08-06T20:34:03.4281480Z  echo "Runner is from Google Cloud Platform, No info on ec2 metadata" 2024-08-06T20:34:03.4282022Z  else 2024-08-06T20:34:03.4282450Z  curl -fsSL "http://169.254.169.254/latest/meta-data/${category}" 2024-08-06T20:34:03.4282970Z  fi 2024-08-06T20:34:03.4283261Z } 2024-08-06T20:34:03.4283595Z echo "ami-id: $(get_ec2_metadata ami-id)" 2024-08-06T20:34:03.4284133Z echo "instance-id: $(get_ec2_metadata instance-id)" 2024-08-06T20:34:03.4284739Z echo "instance-type: $(get_ec2_metadata instance-type)" 2024-08-06T20:34:03.4285255Z echo "system info $(uname -a)" 2024-08-06T20:34:03.4292444Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:34:03.4292927Z env: 2024-08-06T20:34:03.4293209Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:34:03.4293561Z ##[endgroup] 2024-08-06T20:34:03.4392732Z ami-id: ami-06c68f701d8090592 2024-08-06T20:34:03.4454208Z instance-id: i-0de4eefb8fab684da 2024-08-06T20:34:03.4513575Z instance-type: g3.4xlarge 2024-08-06T20:34:03.4525806Z system info Linux ip-10-0-62-112.ec2.internal 6.1.94-99.176.amzn2023.x86_64 #1 SMP PREEMPT_DYNAMIC Tue Jun 18 14:57:56 UTC 2024 x86_64 x86_64 x86_64 GNU/Linux 2024-08-06T20:34:03.4581396Z ##[group]Run echo "IN_ARC_RUNNER=$([ -f /.inarc ] && echo true || echo false)" >> $GITHUB_OUTPUT 2024-08-06T20:34:03.4582209Z echo "IN_ARC_RUNNER=$([ -f /.inarc ] && echo true || echo false)" >> $GITHUB_OUTPUT 2024-08-06T20:34:03.4588644Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:34:03.4589129Z env: 2024-08-06T20:34:03.4589421Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:34:03.4589754Z ##[endgroup] 2024-08-06T20:34:03.4687888Z ##[group]Run if systemctl is-active --quiet docker; then 2024-08-06T20:34:03.4688469Z if systemctl is-active --quiet docker; then 2024-08-06T20:34:03.4688965Z  echo "Docker daemon is running..."; 2024-08-06T20:34:03.4689368Z else 2024-08-06T20:34:03.4689835Z  echo "Starting docker deamon..." && sudo systemctl start docker; 2024-08-06T20:34:03.4690367Z fi 2024-08-06T20:34:03.4696622Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:34:03.4697106Z env: 2024-08-06T20:34:03.4697398Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:34:03.4697729Z ##[endgroup] 2024-08-06T20:34:03.4785449Z Docker daemon is running... 2024-08-06T20:34:03.4870126Z ##[group]Run nick-fields/retry@3e91a01664abd3c5cd539100d10d33b9c5b68482 2024-08-06T20:34:03.4870654Z with: 2024-08-06T20:34:03.4870936Z shell: bash 2024-08-06T20:34:03.4871237Z timeout_minutes: 5 2024-08-06T20:34:03.4871542Z max_attempts: 3 2024-08-06T20:34:03.4871856Z retry_wait_seconds: 30 2024-08-06T20:34:03.4873396Z command: AWS_ACCOUNT_ID=$(aws sts get-caller-identity|grep Account|cut -f4 -d\") aws ecr get-login-password --region "$AWS_DEFAULT_REGION" | docker login --username AWS \ --password-stdin "$AWS_ACCOUNT_ID.dkr.ecr.$AWS_DEFAULT_REGION.amazonaws.com" 2024-08-06T20:34:03.4875515Z polling_interval_seconds: 1 2024-08-06T20:34:03.4876165Z warning_on_retry: true 2024-08-06T20:34:03.4876523Z continue_on_error: false 2024-08-06T20:34:03.4876868Z env: 2024-08-06T20:34:03.4877133Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:34:03.4877482Z AWS_RETRY_MODE: standard 2024-08-06T20:34:03.4877828Z AWS_MAX_ATTEMPTS: 5 2024-08-06T20:34:03.4878248Z AWS_DEFAULT_REGION: us-east-1 2024-08-06T20:34:03.4878868Z ##[endgroup] 2024-08-06T20:34:04.9478750Z WARNING! Your password will be stored unencrypted in /home/ec2-user/.docker/config.json. 2024-08-06T20:34:04.9479529Z Configure a credential helper to remove this warning. See 2024-08-06T20:34:04.9480250Z https://docs.docker.com/engine/reference/commandline/login/#credentials-store 2024-08-06T20:34:04.9480749Z 2024-08-06T20:34:04.9481038Z Login Succeeded 2024-08-06T20:34:05.5635300Z Command completed after 1 attempt(s). 2024-08-06T20:34:05.5716285Z ##[group]Run env | grep '^GITHUB' >> "/tmp/github_env_${GITHUB_RUN_ID}" 2024-08-06T20:34:05.5717027Z env | grep '^GITHUB' >> "/tmp/github_env_${GITHUB_RUN_ID}" 2024-08-06T20:34:05.5717650Z env | grep '^CI' >> "/tmp/github_env_${GITHUB_RUN_ID}" 2024-08-06T20:34:05.5725343Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:34:05.5725855Z env: 2024-08-06T20:34:05.5726138Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:34:05.5726504Z ##[endgroup] 2024-08-06T20:34:05.5830732Z ##[group]Run # ignore expansion of "docker ps -q" since it could be empty 2024-08-06T20:34:05.5831444Z # ignore expansion of "docker ps -q" since it could be empty 2024-08-06T20:34:05.5831972Z # shellcheck disable=SC2046 2024-08-06T20:34:05.5832413Z docker stop $(docker ps -q) || true 2024-08-06T20:34:05.5832872Z # Prune all of the docker images 2024-08-06T20:34:05.5833281Z docker system prune -af 2024-08-06T20:34:05.5839990Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:34:05.5840476Z env: 2024-08-06T20:34:05.5840762Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:34:05.5841110Z ##[endgroup] 2024-08-06T20:34:05.6175808Z "docker stop" requires at least 1 argument. 2024-08-06T20:34:05.6176306Z See 'docker stop --help'. 2024-08-06T20:34:05.6176539Z 2024-08-06T20:34:05.6176754Z Usage: docker stop [OPTIONS] CONTAINER [CONTAINER...] 2024-08-06T20:34:05.6177118Z 2024-08-06T20:34:05.6177270Z Stop one or more running containers 2024-08-06T20:34:05.6371042Z Total reclaimed space: 0B 2024-08-06T20:34:05.6419764Z ##[group]Run set +e 2024-08-06T20:34:05.6420158Z set +e 2024-08-06T20:34:05.6420497Z set -x 2024-08-06T20:34:05.6420804Z  2024-08-06T20:34:05.6421149Z PT_DOMAIN=download.pytorch.org 2024-08-06T20:34:05.6421947Z # TODO: Flaky access to download.pytorch.org https://github.com/pytorch/pytorch/issues/100400, 2024-08-06T20:34:05.6422972Z # cleaning this up once the issue is fixed. There are more than one resolved IP here, the last 2024-08-06T20:34:05.6423731Z # one is returned at random 2024-08-06T20:34:05.6424287Z RESOLVED_IP=$(dig -4 +short "${PT_DOMAIN}" | tail -n1) 2024-08-06T20:34:05.6424800Z  2024-08-06T20:34:05.6425143Z if [ -z "${RESOLVED_IP}" ]; then 2024-08-06T20:34:05.6425752Z  echo "Couldn't resolve ${PT_DOMAIN}, retrying with Google DNS..." 2024-08-06T20:34:05.6426465Z  RESOLVED_IP=$(dig -4 +short "${PT_DOMAIN}" @8.8.8.8 | tail -n1) 2024-08-06T20:34:05.6427024Z  2024-08-06T20:34:05.6427366Z  if [ -z "${RESOLVED_IP}" ]; then 2024-08-06T20:34:05.6427915Z  echo "Couldn't resolve ${PT_DOMAIN}, exiting..." 2024-08-06T20:34:05.6428421Z  exit 1 2024-08-06T20:34:05.6428759Z  fi 2024-08-06T20:34:05.6429266Z fi 2024-08-06T20:34:05.6429597Z  2024-08-06T20:34:05.6429980Z if grep -r "${PT_DOMAIN}" /etc/hosts; then 2024-08-06T20:34:05.6430491Z  # Clean up any old records first 2024-08-06T20:34:05.6431156Z  sudo sed -i "/${PT_DOMAIN}/d" /etc/hosts 2024-08-06T20:34:05.6431625Z fi 2024-08-06T20:34:05.6431916Z  2024-08-06T20:34:05.6432359Z echo "${RESOLVED_IP} ${PT_DOMAIN}" | sudo tee -a /etc/hosts 2024-08-06T20:34:05.6432916Z cat /etc/hosts 2024-08-06T20:34:05.6439960Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:34:05.6440483Z env: 2024-08-06T20:34:05.6440787Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:34:05.6441144Z ##[endgroup] 2024-08-06T20:34:05.6468537Z + PT_DOMAIN=download.pytorch.org 2024-08-06T20:34:05.6474766Z ++ dig -4 +short download.pytorch.org 2024-08-06T20:34:05.6475694Z ++ tail -n1 2024-08-06T20:34:05.6997946Z + RESOLVED_IP=18.160.10.22 2024-08-06T20:34:05.6998726Z + '[' -z 18.160.10.22 ']' 2024-08-06T20:34:05.6999486Z + grep -r download.pytorch.org /etc/hosts 2024-08-06T20:34:05.7011807Z 108.138.85.2 download.pytorch.org 2024-08-06T20:34:05.7012827Z + sudo sed -i /download.pytorch.org/d /etc/hosts 2024-08-06T20:34:05.8506959Z + echo '18.160.10.22 download.pytorch.org' 2024-08-06T20:34:05.8507617Z + sudo tee -a /etc/hosts 2024-08-06T20:34:05.9034803Z 18.160.10.22 download.pytorch.org 2024-08-06T20:34:05.9053481Z + cat /etc/hosts 2024-08-06T20:34:05.9062938Z 127.0.0.1 localhost localhost.localdomain localhost4 localhost4.localdomain4 2024-08-06T20:34:05.9069475Z ::1 localhost6 localhost6.localdomain6 2024-08-06T20:34:05.9069997Z 18.160.10.22 download.pytorch.org 2024-08-06T20:34:05.9231449Z ##[group]Run pytorch/test-infra/.github/actions/calculate-docker-image@main 2024-08-06T20:34:05.9232087Z with: 2024-08-06T20:34:05.9232994Z docker-image-name: 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:05.9233996Z docker-build-dir: .ci/docker 2024-08-06T20:34:05.9234374Z working-directory: . 2024-08-06T20:34:05.9234827Z docker-registry: 308535385114.dkr.ecr.us-east-1.amazonaws.com 2024-08-06T20:34:05.9235335Z force-push: false 2024-08-06T20:34:05.9235654Z env: 2024-08-06T20:34:05.9235941Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:34:05.9236271Z ##[endgroup] 2024-08-06T20:34:05.9269856Z ##[group]Run set -ex 2024-08-06T20:34:05.9270260Z set -ex 2024-08-06T20:34:05.9270567Z  2024-08-06T20:34:05.9271066Z # If the docker build directory or the build script doesn't exist, the action will 2024-08-06T20:34:05.9271939Z # gracefully return the docker image name as it is. Pulling docker image in Linux 2024-08-06T20:34:05.9272637Z # job could then download the pre-built image as usual 2024-08-06T20:34:05.9273283Z if [[ ! -d "${DOCKER_BUILD_DIR}" ]] || [[ ! -f "${DOCKER_BUILD_DIR}/build.sh" ]]; then 2024-08-06T20:34:05.9273878Z  echo "skip=true" >> "${GITHUB_OUTPUT}" 2024-08-06T20:34:05.9274445Z  echo "docker-image=${DOCKER_IMAGE_NAME}" >> "${GITHUB_OUTPUT}" 2024-08-06T20:34:05.9274950Z  2024-08-06T20:34:05.9275418Z  echo "There is no Docker build script in ${REPO_NAME} repo, skipping..." 2024-08-06T20:34:05.9275991Z  exit 0 2024-08-06T20:34:05.9276282Z else 2024-08-06T20:34:05.9276636Z  echo "skip=false" >> "${GITHUB_OUTPUT}" 2024-08-06T20:34:05.9277059Z fi 2024-08-06T20:34:05.9277329Z  2024-08-06T20:34:05.9277764Z if [[ "${DOCKER_IMAGE_NAME}" == *"${DOCKER_REGISTRY}/${REPO_NAME}"* ]]; then 2024-08-06T20:34:05.9278515Z  # The docker image name already includes the ECR prefix and tag, so we can just 2024-08-06T20:34:05.9279169Z  # use it as it is, but first let's extract the tag 2024-08-06T20:34:05.9279779Z  DOCKER_TAG=$(echo "${DOCKER_IMAGE_NAME}" | awk -F '[:,]' '{print $2}') 2024-08-06T20:34:05.9280414Z  echo "docker-tag=${DOCKER_TAG}" >> "${GITHUB_OUTPUT}" 2024-08-06T20:34:05.9281018Z  echo "docker-image=${DOCKER_IMAGE_NAME}" >> "${GITHUB_OUTPUT}" 2024-08-06T20:34:05.9281753Z else 2024-08-06T20:34:05.9282166Z  DOCKER_TAG=$(git rev-parse HEAD:"${DOCKER_BUILD_DIR}") 2024-08-06T20:34:05.9282742Z  echo "docker-tag=${DOCKER_TAG}" >> "${GITHUB_OUTPUT}" 2024-08-06T20:34:05.9283531Z  echo "docker-image=${DOCKER_REGISTRY}/${REPO_NAME}/${DOCKER_IMAGE_NAME}:${DOCKER_TAG}" >> "${GITHUB_OUTPUT}" 2024-08-06T20:34:05.9284223Z fi 2024-08-06T20:34:05.9291872Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:34:05.9292342Z env: 2024-08-06T20:34:05.9292637Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:34:05.9292977Z REPO_NAME: pytorch 2024-08-06T20:34:05.9293909Z DOCKER_IMAGE_NAME: 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:05.9294919Z DOCKER_BUILD_DIR: .ci/docker 2024-08-06T20:34:05.9295395Z DOCKER_REGISTRY: 308535385114.dkr.ecr.us-east-1.amazonaws.com 2024-08-06T20:34:05.9295869Z ##[endgroup] 2024-08-06T20:34:05.9323789Z + [[ ! -d .ci/docker ]] 2024-08-06T20:34:05.9324205Z + [[ ! -f .ci/docker/build.sh ]] 2024-08-06T20:34:05.9324584Z + echo skip=false 2024-08-06T20:34:05.9325776Z + [[ 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 == *\3\0\8\5\3\5\3\8\5\1\1\4\.\d\k\r\.\e\c\r\.\u\s\-\e\a\s\t\-\1\.\a\m\a\z\o\n\a\w\s\.\c\o\m\/\p\y\t\o\r\c\h* ]] 2024-08-06T20:34:05.9331914Z ++ echo 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:05.9333113Z ++ awk -F '[:,]' '{print $2}' 2024-08-06T20:34:05.9355311Z + DOCKER_TAG=02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:05.9355887Z + echo docker-tag=02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:05.9357220Z + echo docker-image=308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:05.9401741Z ##[group]Run set +e 2024-08-06T20:34:05.9402156Z set +e 2024-08-06T20:34:05.9402452Z set -x 2024-08-06T20:34:05.9402755Z  2024-08-06T20:34:05.9403042Z login() { 2024-08-06T20:34:05.9403629Z  aws ecr get-login-password --region us-east-1 | docker login -u AWS --password-stdin "$1" 2024-08-06T20:34:05.9404795Z } 2024-08-06T20:34:05.9405093Z  2024-08-06T20:34:05.9405393Z retry () { 2024-08-06T20:34:05.9405746Z  $* || (sleep 1 && $*) || (sleep 2 && $*) 2024-08-06T20:34:05.9406156Z } 2024-08-06T20:34:05.9406446Z  2024-08-06T20:34:05.9406745Z retry login "${DOCKER_REGISTRY}" 2024-08-06T20:34:05.9407148Z  2024-08-06T20:34:05.9407577Z # Check if image already exists, if it does then skip building it 2024-08-06T20:34:05.9408183Z if docker manifest inspect "${DOCKER_IMAGE}"; then 2024-08-06T20:34:05.9408662Z  exit 0 2024-08-06T20:34:05.9408963Z fi 2024-08-06T20:34:05.9409228Z  2024-08-06T20:34:05.9409685Z # NB: This part requires a full checkout. Otherwise, the merge base will 2024-08-06T20:34:05.9410397Z # be empty. The default action would be to continue rebuild the image 2024-08-06T20:34:05.9411025Z if [[ "$BASE_REVISION" = "$(git rev-parse HEAD)" ]]; then 2024-08-06T20:34:05.9411606Z  # if we're on the base branch then use the parent commit 2024-08-06T20:34:05.9412120Z  MERGE_BASE=$(git rev-parse HEAD~) 2024-08-06T20:34:05.9412518Z else 2024-08-06T20:34:05.9412946Z  # otherwise we're on a PR, so use the most recent base commit 2024-08-06T20:34:05.9413546Z  MERGE_BASE=$(git merge-base HEAD "$BASE_REVISION") 2024-08-06T20:34:05.9414016Z fi 2024-08-06T20:34:05.9414287Z  2024-08-06T20:34:05.9414601Z if [[ -z "${MERGE_BASE}" ]]; then 2024-08-06T20:34:05.9415287Z  echo "rebuild=true" >> "${GITHUB_OUTPUT}" 2024-08-06T20:34:05.9415710Z  2024-08-06T20:34:05.9416298Z  echo "Finding merge base only works with full checkout, please set fetch-depth to 0, continuing ..." 2024-08-06T20:34:05.9416979Z  exit 0 2024-08-06T20:34:05.9417269Z fi 2024-08-06T20:34:05.9417555Z  2024-08-06T20:34:05.9417952Z if ! git rev-parse "${MERGE_BASE}:${DOCKER_BUILD_DIR}"; then 2024-08-06T20:34:05.9418786Z  echo "Directory '${DOCKER_BUILD_DIR}' not found in commit $MERGE_BASE, you should rebase onto a more recent commit" 2024-08-06T20:34:05.9419518Z  exit 1 2024-08-06T20:34:05.9419822Z fi 2024-08-06T20:34:05.9420091Z  2024-08-06T20:34:05.9420555Z PREVIOUS_DOCKER_TAG=$(git rev-parse "${MERGE_BASE}:${DOCKER_BUILD_DIR}") 2024-08-06T20:34:05.9421366Z # If no image exists but the hash is the same as the previous hash then we should error out here 2024-08-06T20:34:05.9422094Z if [[ "${PREVIOUS_DOCKER_TAG}" == "${DOCKER_TAG}" ]]; then 2024-08-06T20:34:05.9422941Z  echo "WARNING: Something has gone wrong and the previous image isn't available for the merge-base of your branch" 2024-08-06T20:34:05.9423879Z  echo " Will re-build docker image to store in local cache, TTS may be longer" 2024-08-06T20:34:05.9424462Z fi 2024-08-06T20:34:05.9424735Z  2024-08-06T20:34:05.9425225Z echo "rebuild=true" >> "${GITHUB_OUTPUT}" 2024-08-06T20:34:05.9431527Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:34:05.9432015Z env: 2024-08-06T20:34:05.9432310Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:34:05.9432660Z DOCKER_BUILD_DIR: .ci/docker 2024-08-06T20:34:05.9433104Z BASE_REVISION: 1736af7cf736184c356be1bb00f59fb2feea6d7d 2024-08-06T20:34:05.9434154Z DOCKER_IMAGE: 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:05.9435222Z DOCKER_TAG: 02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:05.9435780Z DOCKER_REGISTRY: 308535385114.dkr.ecr.us-east-1.amazonaws.com 2024-08-06T20:34:05.9436282Z ##[endgroup] 2024-08-06T20:34:05.9462459Z + retry login 308535385114.dkr.ecr.us-east-1.amazonaws.com 2024-08-06T20:34:05.9463045Z + login 308535385114.dkr.ecr.us-east-1.amazonaws.com 2024-08-06T20:34:05.9465413Z + aws ecr get-login-password --region us-east-1 2024-08-06T20:34:05.9466572Z + docker login -u AWS --password-stdin 308535385114.dkr.ecr.us-east-1.amazonaws.com 2024-08-06T20:34:06.6072598Z WARNING! Your password will be stored unencrypted in /home/ec2-user/.docker/config.json. 2024-08-06T20:34:06.6073358Z Configure a credential helper to remove this warning. See 2024-08-06T20:34:06.6074077Z https://docs.docker.com/engine/reference/commandline/login/#credentials-store 2024-08-06T20:34:06.6074564Z 2024-08-06T20:34:06.6074831Z Login Succeeded 2024-08-06T20:34:06.6089183Z + docker manifest inspect 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:06.8518154Z { 2024-08-06T20:34:06.8518654Z "schemaVersion": 2, 2024-08-06T20:34:06.8519214Z "mediaType": "application/vnd.docker.distribution.manifest.v2+json", 2024-08-06T20:34:06.8519776Z "config": { 2024-08-06T20:34:06.8520263Z "mediaType": "application/vnd.docker.container.image.v1+json", 2024-08-06T20:34:06.8520825Z "size": 48439, 2024-08-06T20:34:06.8521380Z "digest": "sha256:6ec36276acd88c9be8b44d856744037d399b35f4bb1703e637c27ae2b254c901" 2024-08-06T20:34:06.8522025Z }, 2024-08-06T20:34:06.8522293Z "layers": [ 2024-08-06T20:34:06.8522598Z { 2024-08-06T20:34:06.8523053Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8523605Z "size": 28580681, 2024-08-06T20:34:06.8524176Z "digest": "sha256:7a2c559011895d255fce249c00396abff5ae7e0c0a92931d0ed493e71de78e3a" 2024-08-06T20:34:06.8525295Z }, 2024-08-06T20:34:06.8525548Z { 2024-08-06T20:34:06.8525985Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8526520Z "size": 7943451, 2024-08-06T20:34:06.8527021Z "digest": "sha256:224fe954d7252f10539d243d6c9688806f7d13ad775ed02e7f7c79077844510d" 2024-08-06T20:34:06.8527607Z }, 2024-08-06T20:34:06.8527872Z { 2024-08-06T20:34:06.8528296Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8528866Z "size": 55728572, 2024-08-06T20:34:06.8529432Z "digest": "sha256:75722010b82e31715876aeeed0b2cee414296f0124fdfa061ab845ba2a158450" 2024-08-06T20:34:06.8530047Z }, 2024-08-06T20:34:06.8530320Z { 2024-08-06T20:34:06.8530772Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8531317Z "size": 186, 2024-08-06T20:34:06.8531878Z "digest": "sha256:d527cbbb87e3016fd72a18a9b468c945ad0ca27c5770b39debd6ed704db3a195" 2024-08-06T20:34:06.8532531Z }, 2024-08-06T20:34:06.8532795Z { 2024-08-06T20:34:06.8533246Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8533811Z "size": 6886, 2024-08-06T20:34:06.8534352Z "digest": "sha256:b57676e46aee1a8c82e528d78e5a13e31142524eea31c8b213d69ddcb6f1fe80" 2024-08-06T20:34:06.8535023Z }, 2024-08-06T20:34:06.8535307Z { 2024-08-06T20:34:06.8535743Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8536605Z + exit 0 2024-08-06T20:34:06.8536926Z "size": 1329001756, 2024-08-06T20:34:06.8537515Z "digest": "sha256:a8c1e85b5e14cec7af70bf304cb4d4cee6a1d25eb8215b2cf4fdc33e5af5e108" 2024-08-06T20:34:06.8538168Z }, 2024-08-06T20:34:06.8538440Z { 2024-08-06T20:34:06.8538870Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8539435Z "size": 62501, 2024-08-06T20:34:06.8540005Z "digest": "sha256:a41a8d1c11c8d80fe4e82b0d05478f8d51176ff20b8350905fc1b25c93a51198" 2024-08-06T20:34:06.8540631Z }, 2024-08-06T20:34:06.8540914Z { 2024-08-06T20:34:06.8541370Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8541921Z "size": 1684, 2024-08-06T20:34:06.8542473Z "digest": "sha256:0c12278907551c2962927d27c115f6f7bf0df894318b8aea6ece3ef01ccd0a8a" 2024-08-06T20:34:06.8543107Z }, 2024-08-06T20:34:06.8543362Z { 2024-08-06T20:34:06.8543814Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8544372Z "size": 1523, 2024-08-06T20:34:06.8544953Z "digest": "sha256:d8d1234baab3ec9ccb8bb710fc6b8ff6c10896ba2e8d27a347583eca770f9ff1" 2024-08-06T20:34:06.8545653Z }, 2024-08-06T20:34:06.8545912Z { 2024-08-06T20:34:06.8546343Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8546880Z "size": 2528295403, 2024-08-06T20:34:06.8547405Z "digest": "sha256:7ed32bc8e4696fcdb2feef850781160597b2275ad756819c4add88236b0577d5" 2024-08-06T20:34:06.8548001Z }, 2024-08-06T20:34:06.8548262Z { 2024-08-06T20:34:06.8548682Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8549245Z "size": 86016, 2024-08-06T20:34:06.8549813Z "digest": "sha256:ec1e7978c1fe161ced1d98092a51e7c5953ca5fda5577f54df9dbda4afff1b2b" 2024-08-06T20:34:06.8550434Z }, 2024-08-06T20:34:06.8550704Z { 2024-08-06T20:34:06.8551154Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8551702Z "size": 1823, 2024-08-06T20:34:06.8552263Z "digest": "sha256:66b43372aa397c4303ca4e0e1122516909bca0c87b9b4bfb3972b8fd0c1d4390" 2024-08-06T20:34:06.8552898Z }, 2024-08-06T20:34:06.8553149Z { 2024-08-06T20:34:06.8553591Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8554157Z "size": 246768020, 2024-08-06T20:34:06.8554713Z "digest": "sha256:b6662193c745ec6b991e800e920c233379c7c0e74f2f64d9b82dd5dc4a27eb14" 2024-08-06T20:34:06.8555341Z }, 2024-08-06T20:34:06.8555608Z { 2024-08-06T20:34:06.8556041Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8556724Z "size": 545, 2024-08-06T20:34:06.8557303Z "digest": "sha256:5be2b638d110dd5ed631ce7ddf7eefa26b3abd49cf3ab845be5ecb3daec46b67" 2024-08-06T20:34:06.8557940Z }, 2024-08-06T20:34:06.8558214Z { 2024-08-06T20:34:06.8558656Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8559206Z "size": 1283, 2024-08-06T20:34:06.8559774Z "digest": "sha256:71ca63790839b9bfa870ee6927d5d7b60aaa1fc65b38d3e8fc42ace8911859ef" 2024-08-06T20:34:06.8560415Z }, 2024-08-06T20:34:06.8560670Z { 2024-08-06T20:34:06.8561118Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8561686Z "size": 484, 2024-08-06T20:34:06.8562228Z "digest": "sha256:8a74804dc4fa9ad5369e1ae6677a4e17bcc2c53d209a67738dbc795420066650" 2024-08-06T20:34:06.8562867Z }, 2024-08-06T20:34:06.8563126Z { 2024-08-06T20:34:06.8563576Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8564153Z "size": 91712377, 2024-08-06T20:34:06.8564714Z "digest": "sha256:3bacb5389b745ab1f7590db3db714e689a99ee0d7c709f907ccd6906f39905c5" 2024-08-06T20:34:06.8565367Z }, 2024-08-06T20:34:06.8565643Z { 2024-08-06T20:34:06.8566079Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8566655Z "size": 3231, 2024-08-06T20:34:06.8567348Z "digest": "sha256:a8911a72541a4ab35894015b7fb1174ea61c59fedc863dfa563324af5d6ae752" 2024-08-06T20:34:06.8568002Z }, 2024-08-06T20:34:06.8568276Z { 2024-08-06T20:34:06.8568723Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8569280Z "size": 1909, 2024-08-06T20:34:06.8569842Z "digest": "sha256:55d020986bb7c1702235b111c4b83d990fa63ce6045c5ac358a026832bbe8550" 2024-08-06T20:34:06.8570469Z }, 2024-08-06T20:34:06.8570722Z { 2024-08-06T20:34:06.8571172Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8571754Z "size": 700, 2024-08-06T20:34:06.8572287Z "digest": "sha256:679e209a81f89d0be588ce19c3f5191f73883a86e44ab7b3653a3be3f267b69e" 2024-08-06T20:34:06.8572925Z }, 2024-08-06T20:34:06.8573195Z { 2024-08-06T20:34:06.8573625Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8574195Z "size": 2785856582, 2024-08-06T20:34:06.8574776Z "digest": "sha256:d4fb7093f54f7f71e63223ef934b9ab258d53922a199ac4736897cdb90df0683" 2024-08-06T20:34:06.8575399Z }, 2024-08-06T20:34:06.8575669Z { 2024-08-06T20:34:06.8576120Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8576675Z "size": 381, 2024-08-06T20:34:06.8577237Z "digest": "sha256:0d8ab4023e81a9284aef759a1b3c759a907d0cbd39361f3ef0ce4f8c3994f882" 2024-08-06T20:34:06.8577869Z }, 2024-08-06T20:34:06.8606015Z { 2024-08-06T20:34:06.8606543Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8607099Z "size": 12876, 2024-08-06T20:34:06.8607756Z "digest": "sha256:bf191f5f5a0a370ba7136fa618cd8cb1eb76e5f82b8c5773a965cdd105515924" 2024-08-06T20:34:06.8608330Z }, 2024-08-06T20:34:06.8608574Z { 2024-08-06T20:34:06.8608983Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8609512Z "size": 803, 2024-08-06T20:34:06.8610047Z "digest": "sha256:14653e4e245feef24e0aabd8a4cd81c24298f800facc0299f797b161da696a1d" 2024-08-06T20:34:06.8610653Z }, 2024-08-06T20:34:06.8610883Z { 2024-08-06T20:34:06.8611298Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8611835Z "size": 106, 2024-08-06T20:34:06.8612355Z "digest": "sha256:8bdbb000c39dd99342429f8a1183bdb36f312b532ea7e47eb7719fea84c669f6" 2024-08-06T20:34:06.8612983Z }, 2024-08-06T20:34:06.8613214Z { 2024-08-06T20:34:06.8613604Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8614135Z "size": 504, 2024-08-06T20:34:06.8614652Z "digest": "sha256:277383b63c0797c1bd9e23c6f38d6ba85e6e321e2dc6b21fcd832f1935f5af87" 2024-08-06T20:34:06.8615491Z }, 2024-08-06T20:34:06.8615728Z { 2024-08-06T20:34:06.8616134Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8616657Z "size": 121477300, 2024-08-06T20:34:06.8617197Z "digest": "sha256:890313244493db7d65ed3f1cf91a94e6e50bbdb4df87b5bb829a1a3236ffaeb3" 2024-08-06T20:34:06.8617798Z }, 2024-08-06T20:34:06.8618023Z { 2024-08-06T20:34:06.8618443Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8618977Z "size": 109, 2024-08-06T20:34:06.8619502Z "digest": "sha256:f1e3cc0f57ee16caa6ffefa72c065dfe99a5d19a3a352342dfa26b63661589a2" 2024-08-06T20:34:06.8620115Z }, 2024-08-06T20:34:06.8620348Z { 2024-08-06T20:34:06.8620751Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8621284Z "size": 491, 2024-08-06T20:34:06.8621813Z "digest": "sha256:c3cbae3fe054ce8c713ed90c42a306ecc164d8256fd73a14ff7b0e088e150b3f" 2024-08-06T20:34:06.8622425Z }, 2024-08-06T20:34:06.8622656Z { 2024-08-06T20:34:06.8623060Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8623585Z "size": 296, 2024-08-06T20:34:06.8624118Z "digest": "sha256:ccc148c4e7590ced33e52f40edecd2d5ec73cb4a42c87dacaf5c5a7a3912c17b" 2024-08-06T20:34:06.8624733Z }, 2024-08-06T20:34:06.8624960Z { 2024-08-06T20:34:06.8625502Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8626055Z "size": 103, 2024-08-06T20:34:06.8626577Z "digest": "sha256:7912f8c8e80ddc0dfc068c1282e6bd0ffd098b02458818c2c7b52a89c41d8335" 2024-08-06T20:34:06.8627170Z }, 2024-08-06T20:34:06.8627397Z { 2024-08-06T20:34:06.8627805Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8628336Z "size": 1473, 2024-08-06T20:34:06.8628850Z "digest": "sha256:d166ebb28213d6d30940b4fb9739863e9200174e7b550a1591e9028b1a039f83" 2024-08-06T20:34:06.8629441Z }, 2024-08-06T20:34:06.8629685Z { 2024-08-06T20:34:06.8630088Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8630620Z "size": 424146463, 2024-08-06T20:34:06.8631154Z "digest": "sha256:63bf315f789a755602aeb163e43e8173bc191c3dabc75e39ab31d0762bacc84f" 2024-08-06T20:34:06.8631749Z }, 2024-08-06T20:34:06.8631981Z { 2024-08-06T20:34:06.8632387Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8632914Z "size": 159, 2024-08-06T20:34:06.8633425Z "digest": "sha256:bdb818f7b2c8404f3e19777a27592349798986185a1f5b539309bbe8ea96e513" 2024-08-06T20:34:06.8634015Z }, 2024-08-06T20:34:06.8634244Z { 2024-08-06T20:34:06.8634652Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8635177Z "size": 566, 2024-08-06T20:34:06.8635693Z "digest": "sha256:89d8aea05b3a5e45fc1c48daf5ac32901006f7804ce5f2104112c2a2136acf28" 2024-08-06T20:34:06.8636305Z }, 2024-08-06T20:34:06.8636537Z { 2024-08-06T20:34:06.8636953Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8637486Z "size": 35874371, 2024-08-06T20:34:06.8638019Z "digest": "sha256:f1122e19f79064bde97285bf17ca6d8abb889972e5d95a463ffd2382145c1f22" 2024-08-06T20:34:06.8638607Z }, 2024-08-06T20:34:06.8638839Z { 2024-08-06T20:34:06.8639250Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8639774Z "size": 104, 2024-08-06T20:34:06.8640283Z "digest": "sha256:13d6ce3185e9912952041a572e2efa85b4544ec540f6050f750093d180a069f6" 2024-08-06T20:34:06.8640867Z }, 2024-08-06T20:34:06.8641093Z { 2024-08-06T20:34:06.8641503Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8642029Z "size": 425, 2024-08-06T20:34:06.8642548Z "digest": "sha256:feb3f80c392d4aef71730a9673030e955ce0e8a5c41f350eb7a00592d6b0dbb3" 2024-08-06T20:34:06.8643152Z }, 2024-08-06T20:34:06.8643379Z { 2024-08-06T20:34:06.8643799Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8644448Z "size": 20262075, 2024-08-06T20:34:06.8644981Z "digest": "sha256:4fe4cdcdfbd890964b8270a9140a5bf255709a21af4401b0428d91a735e8ac12" 2024-08-06T20:34:06.8645582Z }, 2024-08-06T20:34:06.8645859Z { 2024-08-06T20:34:06.8646245Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8646743Z "size": 440, 2024-08-06T20:34:06.8647242Z "digest": "sha256:be10b99d8ac8cfa04842a726627b1bdc764d3b6f1c591dca7933b86c93208c66" 2024-08-06T20:34:06.8647797Z }, 2024-08-06T20:34:06.8648024Z { 2024-08-06T20:34:06.8648416Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8648937Z "size": 700, 2024-08-06T20:34:06.8649451Z "digest": "sha256:679e209a81f89d0be588ce19c3f5191f73883a86e44ab7b3653a3be3f267b69e" 2024-08-06T20:34:06.8650049Z }, 2024-08-06T20:34:06.8650277Z { 2024-08-06T20:34:06.8650685Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8651224Z "size": 143, 2024-08-06T20:34:06.8651742Z "digest": "sha256:5980a36dfe02695abaecfad21a248c6c1902b07b2c9b69c61c39e342994e2f91" 2024-08-06T20:34:06.8652350Z }, 2024-08-06T20:34:06.8652584Z { 2024-08-06T20:34:06.8652992Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8653521Z "size": 135, 2024-08-06T20:34:06.8654144Z "digest": "sha256:94a4e0b3f19a399451a5f3cc7ddbde73ea16a7f180f7f047bf3ad868072c173f" 2024-08-06T20:34:06.8654758Z }, 2024-08-06T20:34:06.8654992Z { 2024-08-06T20:34:06.8655402Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8655927Z "size": 32, 2024-08-06T20:34:06.8656450Z "digest": "sha256:4f4fb700ef54461cfa02571ae0db9a0dc1e0cdb5577484a6d75e68dc38e8acc1" 2024-08-06T20:34:06.8657054Z }, 2024-08-06T20:34:06.8657279Z { 2024-08-06T20:34:06.8657682Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8658204Z "size": 189, 2024-08-06T20:34:06.8658734Z "digest": "sha256:2012c603f15449503b4671093a9ba6aff4fc99cf4923a92bb446fde7e52d59ee" 2024-08-06T20:34:06.8659328Z }, 2024-08-06T20:34:06.8659554Z { 2024-08-06T20:34:06.8659964Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8660490Z "size": 563, 2024-08-06T20:34:06.8661002Z "digest": "sha256:060890aa9610c5ec0050f85cafaa1f010ff178e2c8b0600aa3c43ad37ed48976" 2024-08-06T20:34:06.8661602Z }, 2024-08-06T20:34:06.8661841Z { 2024-08-06T20:34:06.8662253Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8662784Z "size": 43163116, 2024-08-06T20:34:06.8663328Z "digest": "sha256:c1a64eb8ee12a08340fb5c5a87dc012ff3074a8b683cc399feaa431de7402abd" 2024-08-06T20:34:06.8663930Z }, 2024-08-06T20:34:06.8664161Z { 2024-08-06T20:34:06.8664573Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8665101Z "size": 106, 2024-08-06T20:34:06.8665627Z "digest": "sha256:ed7686d06f1d744c9ec6dd0d75ae1581baefd7809deef8aefa11d54945c7888f" 2024-08-06T20:34:06.8666241Z }, 2024-08-06T20:34:06.8666469Z { 2024-08-06T20:34:06.8666878Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8667439Z "size": 1212, 2024-08-06T20:34:06.8667966Z "digest": "sha256:5c40be0141236773ddf2a3127f247bcc22540d4bebf4f3cc1df53f16f629ee35" 2024-08-06T20:34:06.8668565Z }, 2024-08-06T20:34:06.8668832Z { 2024-08-06T20:34:06.8669263Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8669831Z "size": 700, 2024-08-06T20:34:06.8670385Z "digest": "sha256:679e209a81f89d0be588ce19c3f5191f73883a86e44ab7b3653a3be3f267b69e" 2024-08-06T20:34:06.8670993Z }, 2024-08-06T20:34:06.8671261Z { 2024-08-06T20:34:06.8671704Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8672252Z "size": 138, 2024-08-06T20:34:06.8672803Z "digest": "sha256:95c1963010edc97c994c12e530dc7e5a5717123dfc4378fe8ecca9dbf79de394" 2024-08-06T20:34:06.8673566Z }, 2024-08-06T20:34:06.8673820Z { 2024-08-06T20:34:06.8674263Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8674827Z "size": 120, 2024-08-06T20:34:06.8675358Z "digest": "sha256:5805001913689846871dcb66b59a8d496e3c78fbf4b46c0c55cb11629af04779" 2024-08-06T20:34:06.8675983Z }, 2024-08-06T20:34:06.8676248Z { 2024-08-06T20:34:06.8676681Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8677258Z "size": 1916657670, 2024-08-06T20:34:06.8677829Z "digest": "sha256:b826637ebc384c2f2efbdc841bf6b8f0ac9b6a85060cab5d171d8ed8d49dd3de" 2024-08-06T20:34:06.8678470Z }, 2024-08-06T20:34:06.8678740Z { 2024-08-06T20:34:06.8679168Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8679736Z "size": 173, 2024-08-06T20:34:06.8680279Z "digest": "sha256:859f9c7a63754c26422062903f2a9991578a62fe2f1a81c9ad0f0e9517ab7387" 2024-08-06T20:34:06.8680902Z }, 2024-08-06T20:34:06.8681167Z { 2024-08-06T20:34:06.8681615Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8682165Z "size": 908, 2024-08-06T20:34:06.8682713Z "digest": "sha256:b89ac1530c4a96d2c4c0626a5202eb9e9a05e0d08517e1a5bf165257505309e8" 2024-08-06T20:34:06.8683351Z }, 2024-08-06T20:34:06.8683605Z { 2024-08-06T20:34:06.8684049Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8684708Z "size": 700, 2024-08-06T20:34:06.8685258Z "digest": "sha256:679e209a81f89d0be588ce19c3f5191f73883a86e44ab7b3653a3be3f267b69e" 2024-08-06T20:34:06.8685890Z }, 2024-08-06T20:34:06.8686163Z { 2024-08-06T20:34:06.8686594Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8687159Z "size": 134, 2024-08-06T20:34:06.8687720Z "digest": "sha256:4f10deed2e003fe5f78af780a6b4a71d0107d3ee59e41de3f35c031ca08e9d4d" 2024-08-06T20:34:06.8688346Z }, 2024-08-06T20:34:06.8688625Z { 2024-08-06T20:34:06.8689072Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8689629Z "size": 32, 2024-08-06T20:34:06.8690189Z "digest": "sha256:4f4fb700ef54461cfa02571ae0db9a0dc1e0cdb5577484a6d75e68dc38e8acc1" 2024-08-06T20:34:06.8690837Z }, 2024-08-06T20:34:06.8691088Z { 2024-08-06T20:34:06.8691532Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8692096Z "size": 156, 2024-08-06T20:34:06.8692627Z "digest": "sha256:336420751f1de11d750328660fdb6ebb9051881d009d399e893c10d61ba69b0c" 2024-08-06T20:34:06.8693255Z }, 2024-08-06T20:34:06.8693522Z { 2024-08-06T20:34:06.8693950Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8694514Z "size": 1841, 2024-08-06T20:34:06.8695059Z "digest": "sha256:f7f49611427c9bdc74d97703f780519d1d7d2b95a5377f6f625c8884cbc21d4e" 2024-08-06T20:34:06.8695663Z }, 2024-08-06T20:34:06.8695932Z { 2024-08-06T20:34:06.8696561Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8697105Z "size": 7529783, 2024-08-06T20:34:06.8697634Z "digest": "sha256:628b460c253a663b1f76b99fd2f00d63872fce39b0830a3b45bdeec4f5244660" 2024-08-06T20:34:06.8698202Z }, 2024-08-06T20:34:06.8698462Z { 2024-08-06T20:34:06.8698892Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8699438Z "size": 164, 2024-08-06T20:34:06.8699999Z "digest": "sha256:98e88ff103238559de2c0c76e43d76a01b94584edee356532b7723d1fd39dd85" 2024-08-06T20:34:06.8700623Z }, 2024-08-06T20:34:06.8700871Z { 2024-08-06T20:34:06.8701319Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8701883Z "size": 7944, 2024-08-06T20:34:06.8702432Z "digest": "sha256:6abf825f7962d4bc769dde6a63a4132694ecb9ba0f17006085d8c339aeedf887" 2024-08-06T20:34:06.8703069Z }, 2024-08-06T20:34:06.8703337Z { 2024-08-06T20:34:06.8703765Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8704908Z "size": 8063, 2024-08-06T20:34:06.8705443Z "digest": "sha256:844414c41546bd3c4dd14a45bbd58cca4a2aa0e8f37a781f8c386736ae4d4081" 2024-08-06T20:34:06.8706018Z }, 2024-08-06T20:34:06.8706270Z { 2024-08-06T20:34:06.8706702Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8707253Z "size": 300, 2024-08-06T20:34:06.8707816Z "digest": "sha256:b92a0d83e22950e600ffd0f6391f5c20b499107ba973cab4d7a54a5c65a922b1" 2024-08-06T20:34:06.8708454Z }, 2024-08-06T20:34:06.8708698Z { 2024-08-06T20:34:06.8709135Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8709693Z "size": 7629841, 2024-08-06T20:34:06.8710225Z "digest": "sha256:56e4340bc9e3886f7c099a66772a040a2d34cf0782746af58b0317af979cdfa3" 2024-08-06T20:34:06.8710844Z }, 2024-08-06T20:34:06.8711111Z { 2024-08-06T20:34:06.8711534Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8712108Z "size": 108, 2024-08-06T20:34:06.8712653Z "digest": "sha256:26f48d882588278c8763af295a6bc7147c492d82c6e4395970856a29fb8d77f0" 2024-08-06T20:34:06.8713251Z }, 2024-08-06T20:34:06.8713516Z { 2024-08-06T20:34:06.8713947Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8714514Z "size": 54145778, 2024-08-06T20:34:06.8715089Z "digest": "sha256:b6fe2821ba25ab984577df156aed9b873699ef0f46b6230d8e9a54f9ee22be1e" 2024-08-06T20:34:06.8715863Z }, 2024-08-06T20:34:06.8716138Z { 2024-08-06T20:34:06.8716585Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8717135Z "size": 473, 2024-08-06T20:34:06.8717698Z "digest": "sha256:fae8722cca7f32933f7a25f1491c31ea9a6df4fc1f9fb2360bd29c79b04f1c56" 2024-08-06T20:34:06.8718349Z }, 2024-08-06T20:34:06.8718597Z { 2024-08-06T20:34:06.8719042Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8719606Z "size": 1374858912, 2024-08-06T20:34:06.8720180Z "digest": "sha256:3c7c25c582fced622823798bd877a7fb903ebd4bfecd93c32e43dbd536bb8202" 2024-08-06T20:34:06.8720814Z }, 2024-08-06T20:34:06.8721085Z { 2024-08-06T20:34:06.8721516Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8722088Z "size": 106, 2024-08-06T20:34:06.8722651Z "digest": "sha256:75a49c2f3f0a99be9760740cce745e1ffd508a15bf5ef08077b2032b4d4d97ce" 2024-08-06T20:34:06.8723276Z }, 2024-08-06T20:34:06.8723553Z { 2024-08-06T20:34:06.8724001Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8724550Z "size": 558, 2024-08-06T20:34:06.8725106Z "digest": "sha256:b32c97699ecde27b65bfbbd8ba207755eb28584f2fc64501f4a320045ae969c8" 2024-08-06T20:34:06.8725741Z }, 2024-08-06T20:34:06.8725991Z { 2024-08-06T20:34:06.8726436Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8727005Z "size": 46248557, 2024-08-06T20:34:06.8727546Z "digest": "sha256:b926a85168171349a0ff57c87aa52b9174d3704512eb2e687184f5552883312a" 2024-08-06T20:34:06.8728197Z }, 2024-08-06T20:34:06.8728463Z { 2024-08-06T20:34:06.8728893Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8729463Z "size": 111, 2024-08-06T20:34:06.8730005Z "digest": "sha256:1c5d35b9a7607fd72af03dc281fe78215973e63c51c1b823a704727c8a0944eb" 2024-08-06T20:34:06.8730623Z }, 2024-08-06T20:34:06.8730890Z { 2024-08-06T20:34:06.8731338Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8731888Z "size": 32, 2024-08-06T20:34:06.8732446Z "digest": "sha256:4f4fb700ef54461cfa02571ae0db9a0dc1e0cdb5577484a6d75e68dc38e8acc1" 2024-08-06T20:34:06.8733080Z }, 2024-08-06T20:34:06.8733327Z { 2024-08-06T20:34:06.8733771Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8734317Z "size": 32, 2024-08-06T20:34:06.8734872Z "digest": "sha256:4f4fb700ef54461cfa02571ae0db9a0dc1e0cdb5577484a6d75e68dc38e8acc1" 2024-08-06T20:34:06.8735660Z }, 2024-08-06T20:34:06.8735911Z { 2024-08-06T20:34:06.8736355Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8736925Z "size": 32, 2024-08-06T20:34:06.8737463Z "digest": "sha256:4f4fb700ef54461cfa02571ae0db9a0dc1e0cdb5577484a6d75e68dc38e8acc1" 2024-08-06T20:34:06.8738098Z }, 2024-08-06T20:34:06.8738363Z { 2024-08-06T20:34:06.8738789Z "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", 2024-08-06T20:34:06.8739351Z "size": 32, 2024-08-06T20:34:06.8739906Z "digest": "sha256:4f4fb700ef54461cfa02571ae0db9a0dc1e0cdb5577484a6d75e68dc38e8acc1" 2024-08-06T20:34:06.8740535Z } 2024-08-06T20:34:06.8740799Z ] 2024-08-06T20:34:06.8741065Z } 2024-08-06T20:34:06.8826711Z ##[group]Run tag=${ECR_DOCKER_IMAGE##*/} 2024-08-06T20:34:06.8827206Z tag=${ECR_DOCKER_IMAGE##*/} 2024-08-06T20:34:06.8827700Z echo "docker pull ghcr.io/pytorch/ci-image:${tag/:/-}" 2024-08-06T20:34:06.8834779Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:34:06.8835245Z env: 2024-08-06T20:34:06.8835544Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:34:06.8836499Z ECR_DOCKER_IMAGE: 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:06.8837481Z ##[endgroup] 2024-08-06T20:34:06.8867972Z docker pull ghcr.io/pytorch/ci-image:pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9-02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:06.8932255Z ##[group]Run pytorch/test-infra/.github/actions/pull-docker-image@main 2024-08-06T20:34:06.8932896Z with: 2024-08-06T20:34:06.8933836Z docker-image: 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:06.8935053Z docker-registry: 308535385114.dkr.ecr.us-east-1.amazonaws.com 2024-08-06T20:34:06.8935580Z env: 2024-08-06T20:34:06.8935858Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:34:06.8936239Z ##[endgroup] 2024-08-06T20:34:06.8968005Z ##[group]Run set -x 2024-08-06T20:34:06.8968358Z set -x 2024-08-06T20:34:06.8968642Z set +e 2024-08-06T20:34:06.8968941Z  2024-08-06T20:34:06.8969229Z login() { 2024-08-06T20:34:06.8969811Z  aws ecr get-login-password --region us-east-1 | docker login -u AWS --password-stdin "$1" 2024-08-06T20:34:06.8970462Z } 2024-08-06T20:34:06.8970747Z  2024-08-06T20:34:06.8971078Z retry () { 2024-08-06T20:34:06.8971428Z  $* || (sleep 1 && $*) || (sleep 2 && $*) 2024-08-06T20:34:06.8971842Z } 2024-08-06T20:34:06.8972122Z  2024-08-06T20:34:06.8972418Z retry login "${DOCKER_REGISTRY}" 2024-08-06T20:34:06.8972819Z  2024-08-06T20:34:06.8973101Z set -e 2024-08-06T20:34:06.8973527Z # ignore output since only exit code is used for conditional 2024-08-06T20:34:06.8974146Z # only pull docker image if it's not available locally 2024-08-06T20:34:06.8974835Z if ! docker inspect --type=image "${DOCKER_IMAGE}" >/dev/null 2>/dev/null; then 2024-08-06T20:34:06.8975447Z  retry docker pull "${DOCKER_IMAGE}" 2024-08-06T20:34:06.8975865Z fi 2024-08-06T20:34:06.8982186Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:34:06.8982651Z env: 2024-08-06T20:34:06.8982941Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:34:06.8983896Z DOCKER_IMAGE: 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:06.8984980Z DOCKER_REGISTRY: 308535385114.dkr.ecr.us-east-1.amazonaws.com 2024-08-06T20:34:06.8985454Z ##[endgroup] 2024-08-06T20:34:06.9011717Z + set +e 2024-08-06T20:34:06.9012112Z + retry login 308535385114.dkr.ecr.us-east-1.amazonaws.com 2024-08-06T20:34:06.9012654Z + login 308535385114.dkr.ecr.us-east-1.amazonaws.com 2024-08-06T20:34:06.9015554Z + aws ecr get-login-password --region us-east-1 2024-08-06T20:34:06.9016839Z + docker login -u AWS --password-stdin 308535385114.dkr.ecr.us-east-1.amazonaws.com 2024-08-06T20:34:07.5658438Z WARNING! Your password will be stored unencrypted in /home/ec2-user/.docker/config.json. 2024-08-06T20:34:07.5659767Z Configure a credential helper to remove this warning. See 2024-08-06T20:34:07.5660881Z https://docs.docker.com/engine/reference/commandline/login/#credentials-store 2024-08-06T20:34:07.5661665Z 2024-08-06T20:34:07.5661870Z Login Succeeded 2024-08-06T20:34:07.5672938Z + set -e 2024-08-06T20:34:07.5673956Z + docker inspect --type=image 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:07.5834187Z + retry docker pull 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:07.5835763Z + docker pull 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:34:07.8323374Z 02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9: Pulling from pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9 2024-08-06T20:34:07.8324452Z 7a2c55901189: Pulling fs layer 2024-08-06T20:34:07.8324987Z 224fe954d725: Pulling fs layer 2024-08-06T20:34:07.8325660Z 75722010b82e: Pulling fs layer 2024-08-06T20:34:07.8326054Z d527cbbb87e3: Pulling fs layer 2024-08-06T20:34:07.8326723Z b57676e46aee: Pulling fs layer 2024-08-06T20:34:07.8327118Z a8c1e85b5e14: Pulling fs layer 2024-08-06T20:34:07.8327495Z a41a8d1c11c8: Pulling fs layer 2024-08-06T20:34:07.8327915Z 0c1227890755: Pulling fs layer 2024-08-06T20:34:07.8328461Z d8d1234baab3: Pulling fs layer 2024-08-06T20:34:07.8328968Z 7ed32bc8e469: Pulling fs layer 2024-08-06T20:34:07.8329611Z ec1e7978c1fe: Pulling fs layer 2024-08-06T20:34:07.8330162Z 66b43372aa39: Pulling fs layer 2024-08-06T20:34:07.8330753Z b6662193c745: Pulling fs layer 2024-08-06T20:34:07.8331184Z 5be2b638d110: Pulling fs layer 2024-08-06T20:34:07.8331552Z b57676e46aee: Waiting 2024-08-06T20:34:07.8332163Z 71ca63790839: Pulling fs layer 2024-08-06T20:34:07.8332854Z 8a74804dc4fa: Pulling fs layer 2024-08-06T20:34:07.8333554Z a8c1e85b5e14: Waiting 2024-08-06T20:34:07.8334170Z a41a8d1c11c8: Waiting 2024-08-06T20:34:07.8334808Z 3bacb5389b74: Pulling fs layer 2024-08-06T20:34:07.8335501Z 0c1227890755: Waiting 2024-08-06T20:34:07.8335915Z a8911a72541a: Pulling fs layer 2024-08-06T20:34:07.8336271Z ec1e7978c1fe: Waiting 2024-08-06T20:34:07.8336602Z 7ed32bc8e469: Waiting 2024-08-06T20:34:07.8336930Z d527cbbb87e3: Waiting 2024-08-06T20:34:07.8337234Z 71ca63790839: Waiting 2024-08-06T20:34:07.8337549Z d8d1234baab3: Waiting 2024-08-06T20:34:07.8337868Z 66b43372aa39: Waiting 2024-08-06T20:34:07.8338175Z 8a74804dc4fa: Waiting 2024-08-06T20:34:07.8338559Z 55d020986bb7: Pulling fs layer 2024-08-06T20:34:07.8338933Z 679e209a81f8: Pulling fs layer 2024-08-06T20:34:07.8339285Z 5be2b638d110: Waiting 2024-08-06T20:34:07.8339624Z d4fb7093f54f: Pulling fs layer 2024-08-06T20:34:07.8339984Z b6662193c745: Waiting 2024-08-06T20:34:07.8340436Z 0d8ab4023e81: Pulling fs layer 2024-08-06T20:34:07.8341121Z 679e209a81f8: Waiting 2024-08-06T20:34:07.8341629Z bf191f5f5a0a: Pulling fs layer 2024-08-06T20:34:07.8341988Z 14653e4e245f: Pulling fs layer 2024-08-06T20:34:07.8342361Z 8bdbb000c39d: Pulling fs layer 2024-08-06T20:34:07.8342732Z 277383b63c07: Pulling fs layer 2024-08-06T20:34:07.8343086Z 890313244493: Pulling fs layer 2024-08-06T20:34:07.8343455Z f1e3cc0f57ee: Pulling fs layer 2024-08-06T20:34:07.8343823Z 0d8ab4023e81: Waiting 2024-08-06T20:34:07.8344140Z c3cbae3fe054: Pulling fs layer 2024-08-06T20:34:07.8344512Z ccc148c4e759: Pulling fs layer 2024-08-06T20:34:07.8344881Z bf191f5f5a0a: Waiting 2024-08-06T20:34:07.8345336Z 7912f8c8e80d: Pulling fs layer 2024-08-06T20:34:07.8346084Z 14653e4e245f: Waiting 2024-08-06T20:34:07.8346666Z d166ebb28213: Pulling fs layer 2024-08-06T20:34:07.8347633Z ccc148c4e759: Waiting 2024-08-06T20:34:07.8348198Z a8911a72541a: Waiting 2024-08-06T20:34:07.8348741Z 7912f8c8e80d: Waiting 2024-08-06T20:34:07.8349297Z 63bf315f789a: Pulling fs layer 2024-08-06T20:34:07.8349970Z bdb818f7b2c8: Pulling fs layer 2024-08-06T20:34:07.8350633Z 89d8aea05b3a: Pulling fs layer 2024-08-06T20:34:07.8351297Z f1122e19f790: Pulling fs layer 2024-08-06T20:34:07.8351913Z 63bf315f789a: Waiting 2024-08-06T20:34:07.8352469Z 13d6ce3185e9: Pulling fs layer 2024-08-06T20:34:07.8353096Z d4fb7093f54f: Waiting 2024-08-06T20:34:07.8353656Z bdb818f7b2c8: Waiting 2024-08-06T20:34:07.8354194Z feb3f80c392d: Pulling fs layer 2024-08-06T20:34:07.8354831Z f1122e19f790: Waiting 2024-08-06T20:34:07.8355390Z 4fe4cdcdfbd8: Pulling fs layer 2024-08-06T20:34:07.8355998Z be10b99d8ac8: Pulling fs layer 2024-08-06T20:34:07.8356629Z feb3f80c392d: Waiting 2024-08-06T20:34:07.8357221Z 5980a36dfe02: Pulling fs layer 2024-08-06T20:34:07.8357818Z 4fe4cdcdfbd8: Waiting 2024-08-06T20:34:07.8358389Z 94a4e0b3f19a: Pulling fs layer 2024-08-06T20:34:07.8359013Z 4f4fb700ef54: Pulling fs layer 2024-08-06T20:34:07.8359572Z 2012c603f154: Pulling fs layer 2024-08-06T20:34:07.8360146Z 94a4e0b3f19a: Waiting 2024-08-06T20:34:07.8360703Z 060890aa9610: Pulling fs layer 2024-08-06T20:34:07.8361323Z c1a64eb8ee12: Pulling fs layer 2024-08-06T20:34:07.8361947Z ed7686d06f1d: Pulling fs layer 2024-08-06T20:34:07.8362546Z 5c40be014123: Pulling fs layer 2024-08-06T20:34:07.8363121Z f1e3cc0f57ee: Waiting 2024-08-06T20:34:07.8363821Z d166ebb28213: Waiting 2024-08-06T20:34:07.8364334Z c3cbae3fe054: Waiting 2024-08-06T20:34:07.8364863Z 95c1963010ed: Pulling fs layer 2024-08-06T20:34:07.8365519Z 580500191368: Pulling fs layer 2024-08-06T20:34:07.8366123Z b826637ebc38: Pulling fs layer 2024-08-06T20:34:07.8366667Z c1a64eb8ee12: Waiting 2024-08-06T20:34:07.8367195Z 859f9c7a6375: Pulling fs layer 2024-08-06T20:34:07.8367848Z b89ac1530c4a: Pulling fs layer 2024-08-06T20:34:07.8368498Z 95c1963010ed: Waiting 2024-08-06T20:34:07.8368836Z 580500191368: Waiting 2024-08-06T20:34:07.8369149Z 4f10deed2e00: Pulling fs layer 2024-08-06T20:34:07.8369505Z b826637ebc38: Waiting 2024-08-06T20:34:07.8369821Z 060890aa9610: Waiting 2024-08-06T20:34:07.8370126Z 336420751f1d: Pulling fs layer 2024-08-06T20:34:07.8370481Z ed7686d06f1d: Waiting 2024-08-06T20:34:07.8370814Z f7f49611427c: Pulling fs layer 2024-08-06T20:34:07.8371154Z 4f10deed2e00: Waiting 2024-08-06T20:34:07.8371482Z 628b460c253a: Pulling fs layer 2024-08-06T20:34:07.8371851Z 98e88ff10323: Pulling fs layer 2024-08-06T20:34:07.8372199Z 4f4fb700ef54: Waiting 2024-08-06T20:34:07.8372522Z 628b460c253a: Waiting 2024-08-06T20:34:07.8372857Z 6abf825f7962: Pulling fs layer 2024-08-06T20:34:07.8373207Z 844414c41546: Pulling fs layer 2024-08-06T20:34:07.8373573Z b92a0d83e229: Pulling fs layer 2024-08-06T20:34:07.8373909Z 5c40be014123: Waiting 2024-08-06T20:34:07.8374240Z 56e4340bc9e3: Pulling fs layer 2024-08-06T20:34:07.8374595Z be10b99d8ac8: Waiting 2024-08-06T20:34:07.8374898Z 6abf825f7962: Waiting 2024-08-06T20:34:07.8375241Z 26f48d882588: Pulling fs layer 2024-08-06T20:34:07.8375611Z b6fe2821ba25: Pulling fs layer 2024-08-06T20:34:07.8375949Z 26f48d882588: Waiting 2024-08-06T20:34:07.8376286Z fae8722cca7f: Pulling fs layer 2024-08-06T20:34:07.8376659Z 3c7c25c582fc: Pulling fs layer 2024-08-06T20:34:07.8377012Z 75a49c2f3f0a: Pulling fs layer 2024-08-06T20:34:07.8377369Z 844414c41546: Waiting 2024-08-06T20:34:07.8377690Z b6fe2821ba25: Waiting 2024-08-06T20:34:07.8378004Z b32c97699ecd: Pulling fs layer 2024-08-06T20:34:07.8378375Z fae8722cca7f: Waiting 2024-08-06T20:34:07.8378706Z b926a8516817: Pulling fs layer 2024-08-06T20:34:07.8379056Z 1c5d35b9a760: Pulling fs layer 2024-08-06T20:34:07.8379417Z 75a49c2f3f0a: Waiting 2024-08-06T20:34:07.8379736Z f7f49611427c: Waiting 2024-08-06T20:34:07.8380033Z b926a8516817: Waiting 2024-08-06T20:34:07.8380345Z 98e88ff10323: Waiting 2024-08-06T20:34:07.8380643Z b92a0d83e229: Waiting 2024-08-06T20:34:07.8380958Z 56e4340bc9e3: Waiting 2024-08-06T20:34:07.8381272Z 859f9c7a6375: Waiting 2024-08-06T20:34:07.8381714Z 3c7c25c582fc: Waiting 2024-08-06T20:34:07.8382047Z 1c5d35b9a760: Waiting 2024-08-06T20:34:07.9779648Z 224fe954d725: Verifying Checksum 2024-08-06T20:34:07.9780068Z 224fe954d725: Download complete 2024-08-06T20:34:08.0792350Z d527cbbb87e3: Download complete 2024-08-06T20:34:08.1745953Z 7a2c55901189: Verifying Checksum 2024-08-06T20:34:08.1746629Z 7a2c55901189: Download complete 2024-08-06T20:34:08.1905102Z b57676e46aee: Verifying Checksum 2024-08-06T20:34:08.1905718Z b57676e46aee: Download complete 2024-08-06T20:34:08.2640773Z a41a8d1c11c8: Verifying Checksum 2024-08-06T20:34:08.2641592Z a41a8d1c11c8: Download complete 2024-08-06T20:34:08.3461969Z 0c1227890755: Verifying Checksum 2024-08-06T20:34:08.3462789Z 0c1227890755: Download complete 2024-08-06T20:34:08.4253140Z d8d1234baab3: Verifying Checksum 2024-08-06T20:34:08.4254015Z d8d1234baab3: Download complete 2024-08-06T20:34:08.4411317Z 75722010b82e: Verifying Checksum 2024-08-06T20:34:08.4412246Z 75722010b82e: Download complete 2024-08-06T20:34:08.5395003Z ec1e7978c1fe: Verifying Checksum 2024-08-06T20:34:08.5395506Z ec1e7978c1fe: Download complete 2024-08-06T20:34:08.6494533Z 66b43372aa39: Download complete 2024-08-06T20:34:09.5318936Z 7a2c55901189: Pull complete 2024-08-06T20:34:09.9054812Z 224fe954d725: Pull complete 2024-08-06T20:34:11.0418344Z 75722010b82e: Pull complete 2024-08-06T20:34:11.0708947Z d527cbbb87e3: Pull complete 2024-08-06T20:34:11.1005168Z b57676e46aee: Pull complete 2024-08-06T20:34:11.2125419Z b6662193c745: Verifying Checksum 2024-08-06T20:34:11.2126307Z b6662193c745: Download complete 2024-08-06T20:34:11.3033638Z 5be2b638d110: Verifying Checksum 2024-08-06T20:34:11.3034047Z 5be2b638d110: Download complete 2024-08-06T20:34:11.3796373Z 71ca63790839: Verifying Checksum 2024-08-06T20:34:11.3796800Z 71ca63790839: Download complete 2024-08-06T20:34:11.4685650Z 8a74804dc4fa: Verifying Checksum 2024-08-06T20:34:11.4686045Z 8a74804dc4fa: Download complete 2024-08-06T20:34:12.4419529Z 3bacb5389b74: Verifying Checksum 2024-08-06T20:34:12.4420413Z 3bacb5389b74: Download complete 2024-08-06T20:34:12.5275006Z a8911a72541a: Verifying Checksum 2024-08-06T20:34:12.5275441Z a8911a72541a: Download complete 2024-08-06T20:34:12.6076115Z 55d020986bb7: Download complete 2024-08-06T20:34:12.6883663Z 679e209a81f8: Download complete 2024-08-06T20:34:21.5204249Z a8c1e85b5e14: Verifying Checksum 2024-08-06T20:34:21.5204930Z a8c1e85b5e14: Download complete 2024-08-06T20:34:21.6360209Z 0d8ab4023e81: Verifying Checksum 2024-08-06T20:34:21.6360700Z 0d8ab4023e81: Download complete 2024-08-06T20:34:21.7303683Z bf191f5f5a0a: Download complete 2024-08-06T20:34:21.8116790Z 14653e4e245f: Download complete 2024-08-06T20:34:21.8866317Z 8bdbb000c39d: Verifying Checksum 2024-08-06T20:34:21.8866790Z 8bdbb000c39d: Download complete 2024-08-06T20:34:21.9609050Z 277383b63c07: Verifying Checksum 2024-08-06T20:34:21.9609565Z 277383b63c07: Download complete 2024-08-06T20:34:23.2897945Z 890313244493: Verifying Checksum 2024-08-06T20:34:23.2898752Z 890313244493: Download complete 2024-08-06T20:34:23.3751117Z f1e3cc0f57ee: Download complete 2024-08-06T20:34:23.4605020Z c3cbae3fe054: Verifying Checksum 2024-08-06T20:34:23.4605862Z c3cbae3fe054: Download complete 2024-08-06T20:34:23.5495303Z ccc148c4e759: Download complete 2024-08-06T20:34:23.6222490Z 7912f8c8e80d: Verifying Checksum 2024-08-06T20:34:23.7032545Z 7912f8c8e80d: Download complete 2024-08-06T20:34:23.7033017Z d166ebb28213: Verifying Checksum 2024-08-06T20:34:23.7033426Z d166ebb28213: Download complete 2024-08-06T20:34:27.9901955Z 63bf315f789a: Verifying Checksum 2024-08-06T20:34:27.9902457Z 63bf315f789a: Download complete 2024-08-06T20:34:28.0769732Z bdb818f7b2c8: Verifying Checksum 2024-08-06T20:34:28.0770529Z bdb818f7b2c8: Download complete 2024-08-06T20:34:28.1590361Z 89d8aea05b3a: Download complete 2024-08-06T20:34:28.5990780Z f1122e19f790: Verifying Checksum 2024-08-06T20:34:28.5991600Z f1122e19f790: Download complete 2024-08-06T20:34:28.6837443Z 13d6ce3185e9: Verifying Checksum 2024-08-06T20:34:28.6838292Z 13d6ce3185e9: Download complete 2024-08-06T20:34:28.7689108Z feb3f80c392d: Verifying Checksum 2024-08-06T20:34:28.7691167Z feb3f80c392d: Download complete 2024-08-06T20:34:29.1455617Z 4fe4cdcdfbd8: Verifying Checksum 2024-08-06T20:34:29.1456460Z 4fe4cdcdfbd8: Download complete 2024-08-06T20:34:29.2313321Z be10b99d8ac8: Verifying Checksum 2024-08-06T20:34:29.2313789Z be10b99d8ac8: Download complete 2024-08-06T20:34:29.3812883Z 94a4e0b3f19a: Download complete 2024-08-06T20:34:29.3937240Z 4f4fb700ef54: Verifying Checksum 2024-08-06T20:34:29.3937692Z 4f4fb700ef54: Download complete 2024-08-06T20:34:29.4768699Z 2012c603f154: Verifying Checksum 2024-08-06T20:34:29.4769346Z 2012c603f154: Download complete 2024-08-06T20:34:29.5392467Z 060890aa9610: Verifying Checksum 2024-08-06T20:34:29.5393354Z 060890aa9610: Download complete 2024-08-06T20:34:30.0976595Z c1a64eb8ee12: Verifying Checksum 2024-08-06T20:34:30.0977076Z c1a64eb8ee12: Download complete 2024-08-06T20:34:30.2094238Z ed7686d06f1d: Download complete 2024-08-06T20:34:30.2956468Z 5c40be014123: Verifying Checksum 2024-08-06T20:34:30.2957265Z 5c40be014123: Download complete 2024-08-06T20:34:30.3724503Z 95c1963010ed: Download complete 2024-08-06T20:34:30.5556433Z 580500191368: Verifying Checksum 2024-08-06T20:34:30.5557026Z 580500191368: Download complete 2024-08-06T20:34:33.7780346Z 7ed32bc8e469: Verifying Checksum 2024-08-06T20:34:33.7780894Z 7ed32bc8e469: Download complete 2024-08-06T20:34:33.8553901Z 859f9c7a6375: Verifying Checksum 2024-08-06T20:34:33.8554722Z 859f9c7a6375: Download complete 2024-08-06T20:34:33.9664802Z b89ac1530c4a: Verifying Checksum 2024-08-06T20:34:33.9665391Z b89ac1530c4a: Download complete 2024-08-06T20:34:34.0544675Z 4f10deed2e00: Verifying Checksum 2024-08-06T20:34:34.0545540Z 4f10deed2e00: Download complete 2024-08-06T20:34:34.1392510Z 336420751f1d: Verifying Checksum 2024-08-06T20:34:34.1393585Z 336420751f1d: Download complete 2024-08-06T20:34:34.2274171Z f7f49611427c: Verifying Checksum 2024-08-06T20:34:34.2275003Z f7f49611427c: Download complete 2024-08-06T20:34:34.4386873Z 628b460c253a: Download complete 2024-08-06T20:34:34.5167765Z 98e88ff10323: Verifying Checksum 2024-08-06T20:34:34.5168377Z 98e88ff10323: Download complete 2024-08-06T20:34:34.6117308Z 6abf825f7962: Download complete 2024-08-06T20:34:34.6939346Z 844414c41546: Verifying Checksum 2024-08-06T20:34:34.6939811Z 844414c41546: Download complete 2024-08-06T20:34:34.7743386Z b92a0d83e229: Verifying Checksum 2024-08-06T20:34:34.9560988Z 56e4340bc9e3: Verifying Checksum 2024-08-06T20:34:34.9561519Z 56e4340bc9e3: Download complete 2024-08-06T20:34:35.0547451Z 26f48d882588: Verifying Checksum 2024-08-06T20:34:35.0547913Z 26f48d882588: Download complete 2024-08-06T20:34:35.6511621Z b6fe2821ba25: Verifying Checksum 2024-08-06T20:34:35.6512364Z b6fe2821ba25: Download complete 2024-08-06T20:34:35.7357527Z fae8722cca7f: Download complete 2024-08-06T20:34:37.6312800Z a8c1e85b5e14: Pull complete 2024-08-06T20:34:38.0811909Z a41a8d1c11c8: Pull complete 2024-08-06T20:34:38.4895998Z 0c1227890755: Pull complete 2024-08-06T20:34:38.9345143Z d8d1234baab3: Pull complete 2024-08-06T20:34:40.6505316Z d4fb7093f54f: Verifying Checksum 2024-08-06T20:34:40.6505776Z d4fb7093f54f: Download complete 2024-08-06T20:34:40.7389658Z 75a49c2f3f0a: Verifying Checksum 2024-08-06T20:34:40.7390143Z 75a49c2f3f0a: Download complete 2024-08-06T20:34:40.8162919Z b32c97699ecd: Verifying Checksum 2024-08-06T20:34:40.8163762Z b32c97699ecd: Download complete 2024-08-06T20:34:41.4305282Z b926a8516817: Verifying Checksum 2024-08-06T20:34:41.4305810Z b926a8516817: Download complete 2024-08-06T20:34:41.5134760Z 1c5d35b9a760: Verifying Checksum 2024-08-06T20:34:41.5135548Z 1c5d35b9a760: Download complete 2024-08-06T20:34:49.5323906Z 3c7c25c582fc: Verifying Checksum 2024-08-06T20:34:49.5324356Z 3c7c25c582fc: Download complete 2024-08-06T20:34:49.7889055Z b826637ebc38: Verifying Checksum 2024-08-06T20:34:49.7889530Z b826637ebc38: Download complete 2024-08-06T20:35:15.3358220Z 7ed32bc8e469: Pull complete 2024-08-06T20:35:15.7917450Z ec1e7978c1fe: Pull complete 2024-08-06T20:35:16.2498554Z 66b43372aa39: Pull complete 2024-08-06T20:35:26.0505284Z b6662193c745: Pull complete 2024-08-06T20:35:26.4985511Z 5be2b638d110: Pull complete 2024-08-06T20:35:26.9513011Z 71ca63790839: Pull complete 2024-08-06T20:35:27.3638557Z 8a74804dc4fa: Pull complete 2024-08-06T20:35:30.6077070Z 3bacb5389b74: Pull complete 2024-08-06T20:35:30.8418337Z a8911a72541a: Pull complete 2024-08-06T20:35:31.1838631Z 55d020986bb7: Pull complete 2024-08-06T20:35:31.6299827Z 679e209a81f8: Pull complete 2024-08-06T20:36:34.1711295Z d4fb7093f54f: Pull complete 2024-08-06T20:36:34.6190276Z 0d8ab4023e81: Pull complete 2024-08-06T20:36:35.1040915Z bf191f5f5a0a: Pull complete 2024-08-06T20:36:35.5608546Z 14653e4e245f: Pull complete 2024-08-06T20:36:35.9877306Z 8bdbb000c39d: Pull complete 2024-08-06T20:36:36.4407974Z 277383b63c07: Pull complete 2024-08-06T20:36:40.9349766Z 890313244493: Pull complete 2024-08-06T20:36:41.2608071Z f1e3cc0f57ee: Pull complete 2024-08-06T20:36:41.5736452Z c3cbae3fe054: Pull complete 2024-08-06T20:36:41.9234853Z ccc148c4e759: Pull complete 2024-08-06T20:36:42.2966422Z 7912f8c8e80d: Pull complete 2024-08-06T20:36:42.5870407Z d166ebb28213: Pull complete 2024-08-06T20:36:53.3127586Z 63bf315f789a: Pull complete 2024-08-06T20:36:53.7342107Z bdb818f7b2c8: Pull complete 2024-08-06T20:36:54.0204539Z 89d8aea05b3a: Pull complete 2024-08-06T20:36:55.2783492Z f1122e19f790: Pull complete 2024-08-06T20:36:55.5105256Z 13d6ce3185e9: Pull complete 2024-08-06T20:36:55.7190928Z feb3f80c392d: Pull complete 2024-08-06T20:36:56.5465416Z 4fe4cdcdfbd8: Pull complete 2024-08-06T20:36:57.0048446Z be10b99d8ac8: Pull complete 2024-08-06T20:36:57.8351910Z 5980a36dfe02: Pull complete 2024-08-06T20:36:58.2638804Z 94a4e0b3f19a: Pull complete 2024-08-06T20:36:58.6900525Z 4f4fb700ef54: Pull complete 2024-08-06T20:36:59.1346219Z 2012c603f154: Pull complete 2024-08-06T20:36:59.5838153Z 060890aa9610: Pull complete 2024-08-06T20:37:02.8619469Z c1a64eb8ee12: Pull complete 2024-08-06T20:37:03.2099886Z ed7686d06f1d: Pull complete 2024-08-06T20:37:03.5101302Z 5c40be014123: Pull complete 2024-08-06T20:37:04.3875057Z 95c1963010ed: Pull complete 2024-08-06T20:37:04.7238154Z 580500191368: Pull complete 2024-08-06T20:37:47.7542647Z b826637ebc38: Pull complete 2024-08-06T20:37:48.2134284Z 859f9c7a6375: Pull complete 2024-08-06T20:37:48.6678260Z b89ac1530c4a: Pull complete 2024-08-06T20:37:49.5793987Z 4f10deed2e00: Pull complete 2024-08-06T20:37:50.4962275Z 336420751f1d: Pull complete 2024-08-06T20:37:50.9468629Z f7f49611427c: Pull complete 2024-08-06T20:37:51.5791904Z 628b460c253a: Pull complete 2024-08-06T20:37:52.0394679Z 98e88ff10323: Pull complete 2024-08-06T20:37:52.4835398Z 6abf825f7962: Pull complete 2024-08-06T20:37:52.9297618Z 844414c41546: Pull complete 2024-08-06T20:37:53.3911900Z b92a0d83e229: Pull complete 2024-08-06T20:37:54.9683399Z 56e4340bc9e3: Pull complete 2024-08-06T20:37:55.2628807Z 26f48d882588: Pull complete 2024-08-06T20:37:58.0016520Z b6fe2821ba25: Pull complete 2024-08-06T20:37:58.4545194Z fae8722cca7f: Pull complete 2024-08-06T20:38:14.2102960Z 3c7c25c582fc: Pull complete 2024-08-06T20:38:14.6550572Z 75a49c2f3f0a: Pull complete 2024-08-06T20:38:15.0404605Z b32c97699ecd: Pull complete 2024-08-06T20:38:16.1621606Z b926a8516817: Pull complete 2024-08-06T20:38:16.5576337Z 1c5d35b9a760: Pull complete 2024-08-06T20:38:18.4463005Z Digest: sha256:00f47b036f588ca5ef8866f8635fabba5a95cdf9ff1adae7d2a674ef1d4076e9 2024-08-06T20:38:18.5166382Z Status: Downloaded newer image for 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:38:18.5280793Z 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:38:18.5346751Z ##[group]Run echo "IN_ARC_RUNNER=$([ -f /.inarc ] && echo true || echo false)" >> "$GITHUB_OUTPUT" 2024-08-06T20:38:18.5347631Z echo "IN_ARC_RUNNER=$([ -f /.inarc ] && echo true || echo false)" >> "$GITHUB_OUTPUT" 2024-08-06T20:38:18.5357658Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:38:18.5358148Z env: 2024-08-06T20:38:18.5358444Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:38:18.5358779Z ##[endgroup] 2024-08-06T20:38:18.5641715Z ##[group]Run pytorch/test-infra/.github/actions/setup-nvidia@main 2024-08-06T20:38:18.5642255Z with: 2024-08-06T20:38:18.5642547Z driver-version: 550.54.15 2024-08-06T20:38:18.5642890Z env: 2024-08-06T20:38:18.5643158Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:38:18.5643499Z ##[endgroup] 2024-08-06T20:38:18.5910693Z ##[group]Run nick-fields/retry@3e91a01664abd3c5cd539100d10d33b9c5b68482 2024-08-06T20:38:18.5911223Z with: 2024-08-06T20:38:18.5911509Z timeout_minutes: 10 2024-08-06T20:38:18.5911818Z max_attempts: 3 2024-08-06T20:38:18.5942144Z command: # Is it disgusting to have a full shell script here in this github action? Sure # But is it the best way to make it so that this action relies on nothing else? Absolutely set -eou pipefail DISTRIBUTION=$(. /etc/os-release;echo $ID$VERSION_ID) DRIVER_FN="NVIDIA-Linux-x86_64-${DRIVER_VERSION}.run" install_nvidia_docker2_amzn2() { ( set -x # Needed for yum-config-manager sudo yum install -y yum-utils if [[ "${DISTRIBUTION}" == "amzn2023" ]] ; then YUM_REPO_URL="https://nvidia.github.io/libnvidia-container/stable/rpm/nvidia-container-toolkit.repo" else # Amazon Linux 2 YUM_REPO_URL="https://nvidia.github.io/nvidia-docker/${DISTRIBUTION}/nvidia-docker.repo" fi sudo yum-config-manager --add-repo "${YUM_REPO_URL}" sudo yum install -y nvidia-docker2 sudo systemctl restart docker ) } install_nvidia_docker2_ubuntu20() { ( set -x # Install nvidia-driver package if not installed status="$(dpkg-query -W --showformat='${db:Status-Status}' nvidia-docker2 2>&1)" if [ ! $? = 0 ] || [ ! "$status" = installed ]; then sudo apt-get install -y nvidia-docker2 sudo systemctl restart docker fi ) } pre_install_nvidia_driver_amzn2() { ( # Purge any nvidia driver installed from RHEL repo sudo yum remove -y nvidia-driver-latest-dkms ) } install_nvidia_driver_common() { ( # Try to gather more information about the runner and its existing NVIDIA driver if any echo "Before installing NVIDIA driver" lspci lsmod modinfo nvidia || true HAS_NVIDIA_DRIVER=0 # Check if NVIDIA driver has already been installed if [ -x "$(command -v nvidia-smi)" ]; then set +e # The driver exists, check its version next. Also check only the first GPU if there are more than one of them # so that the same driver version is not print over multiple lines INSTALLED_DRIVER_VERSION=$(nvidia-smi --query-gpu=driver_version --format=csv,noheader --id=0) NVIDIA_SMI_STATUS=$? if [ "$NVIDIA_SMI_STATUS" -ne 0 ] && [ "$NVIDIA_SMI_STATUS" -ne 14 ]; then echo "Failed to get NVIDIA driver version ($INSTALLED_DRIVER_VERSION). Continuing" elif [ "$INSTALLED_DRIVER_VERSION" != "$DRIVER_VERSION" ]; then echo "NVIDIA driver ($INSTALLED_DRIVER_VERSION) has been installed, but we expect to have $DRIVER_VERSION instead. Continuing" else HAS_NVIDIA_DRIVER=1 echo "NVIDIA driver ($INSTALLED_DRIVER_VERSION) has already been installed. Skipping NVIDIA driver installation" fi set -e fi if [ "$HAS_NVIDIA_DRIVER" -eq 0 ]; then # CAUTION: this may need to be updated in future if [ "${DISTRIBUTION}" != ubuntu20.04 ]; then sudo yum groupinstall -y "Development Tools" # ensure our kernel install is the same as our underlying kernel, # groupinstall "Development Tools" has a habit of mismatching kernel headers sudo yum install -y "kernel-devel-uname-r == $(uname -r)" sudo modprobe backlight fi sudo curl -fsL -o /tmp/nvidia_driver "https://s3.amazonaws.com/ossci-linux/nvidia_driver/$DRIVER_FN" set +e sudo /bin/bash /tmp/nvidia_driver -s --no-drm NVIDIA_INSTALLATION_STATUS=$? RESET_GPU=0 if [ "$NVIDIA_INSTALLATION_STATUS" -ne 0 ]; then sudo cat /var/log/nvidia-installer.log # Fail to install NVIDIA driver, try to reset the GPU RESET_GPU=1 elif [ -x "$(command -v nvidia-smi)" ]; then # Check again if nvidia-smi works even if the driver installation completes successfully INSTALLED_DRIVER_VERSION=$(nvidia-smi --query-gpu=driver_version --format=csv,noheader --id=0) NVIDIA_SMI_STATUS=$? if [ "$NVIDIA_SMI_STATUS" -ne 0 ] && [ "$NVIDIA_SMI_STATUS" -ne 14 ]; then RESET_GPU=1 fi fi if [ "$RESET_GPU" -eq 1 ]; then NVIDIA_DEVICES=$(lspci -D | grep -i NVIDIA | cut -d' ' -f1) # The GPU can get stuck in a failure state if somehow the test crashs the GPU microcode. When this # happens, we'll try to reset all NVIDIA devices https://github.com/pytorch/pytorch/issues/88388 for PCI_ID in $NVIDIA_DEVICES; do DEVICE_ENABLED=$(cat /sys/bus/pci/devices/$PCI_ID/enable) echo "Reseting $PCI_ID (enabled state: $DEVICE_ENABLED)" # This requires sudo permission of course echo "1" | sudo tee /sys/bus/pci/devices/$PCI_ID/reset sleep 1 done fi sudo rm -fv /tmp/nvidia_driver set -e fi ) } post_install_nvidia_driver_common() { ( sudo modprobe nvidia || true echo "After installing NVIDIA driver" lspci lsmod modinfo nvidia || true ( set +e nvidia-smi # NB: Annoyingly, nvidia-smi command returns successfully with return code 0 even in # the case where the driver has already crashed as it still can get the driver version # and some basic information like the bus ID. However, the rest of the information # would be missing (ERR!), for example: # # +-----------------------------------------------------------------------------+ # | NVIDIA-SMI 525.89.02 Driver Version: 525.89.02 CUDA Version: 12.0 | # |-------------------------------+----------------------+----------------------+ # | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC | # | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. | # | | | MIG M. | # |===============================+======================+======================| # | 0 ERR! Off | 00000000:00:1E.0 Off | ERR! | # |ERR! ERR! ERR! ERR! / ERR! | 4184MiB / 23028MiB | ERR! Default | # | | | ERR! | # +-------------------------------+----------------------+----------------------+ # # +-----------------------------------------------------------------------------+ # | Processes: | # | GPU GI CI PID Type Process name GPU Memory | # | ID ID Usage | # |=============================================================================| # +-----------------------------------------------------------------------------+ # # This should be reported as a failure instead as it will guarantee to fail when # Docker tries to run with --gpus all # # So, the correct check here is to query one of the missing piece of info like # GPU name, so that the command can fail accordingly nvidia-smi --query-gpu=gpu_name --format=csv,noheader --id=0 NVIDIA_SMI_STATUS=$? # Allowable exit statuses for nvidia-smi, see: https://github.com/NVIDIA/gpu-operator/issues/285 if [ "$NVIDIA_SMI_STATUS" -eq 0 ] || [ "$NVIDIA_SMI_STATUS" -eq 14 ]; then echo "INFO: Ignoring allowed status ${NVIDIA_SMI_STATUS}" else echo "ERROR: nvidia-smi exited with unresolved status ${NVIDIA_SMI_STATUS}" exit ${NVIDIA_SMI_STATUS} fi set -e ) ) } install_nvidia_driver_amzn2() { ( set -x pre_install_nvidia_driver_amzn2 install_nvidia_driver_common post_install_nvidia_driver_common ) } install_nvidia_driver_ubuntu20() { ( set -x install_nvidia_driver_common post_install_nvidia_driver_common ) } echo "== Installing nvidia driver ${DRIVER_FN} ==" case "${DISTRIBUTION}" in amzn*) install_nvidia_driver_amzn2 ;; ubuntu20.04) install_nvidia_driver_ubuntu20 ;; *) echo "ERROR: Unknown distribution ${DISTRIBUTION}" exit 1 ;; esac # Install container toolkit based on distribution echo "== Installing nvidia container toolkit for ${DISTRIBUTION} ==" case "${DISTRIBUTION}" in amzn*) install_nvidia_docker2_amzn2 ;; ubuntu20.04) install_nvidia_docker2_ubuntu20 ;; *) echo "ERROR: Unknown distribution ${DISTRIBUTION}" exit 1 ;; esac echo "GPU_FLAG=--gpus all -e NVIDIA_DRIVER_CAPABILITIES=all" >> "${GITHUB_ENV}" # Fix https://github.com/NVIDIA/nvidia-docker/issues/1648 on runners with # more than one GPUs. This just needs to be run once. The command fails # on subsequent runs and complains that the mode is already on, but that's # ok sudo nvidia-persistenced || true # This should show persistence mode ON nvidia-smi 2024-08-06T20:38:18.5972618Z retry_wait_seconds: 10 2024-08-06T20:38:18.5972988Z polling_interval_seconds: 1 2024-08-06T20:38:18.5973341Z warning_on_retry: true 2024-08-06T20:38:18.5973689Z continue_on_error: false 2024-08-06T20:38:18.5974028Z env: 2024-08-06T20:38:18.5974294Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:38:18.5974643Z DRIVER_VERSION: 550.54.15 2024-08-06T20:38:18.5974967Z ##[endgroup] 2024-08-06T20:38:18.6944654Z == Installing nvidia driver NVIDIA-Linux-x86_64-550.54.15.run == 2024-08-06T20:38:18.6947186Z + pre_install_nvidia_driver_amzn2 2024-08-06T20:38:18.6949713Z + sudo yum remove -y nvidia-driver-latest-dkms 2024-08-06T20:38:19.1402832Z No match for argument: nvidia-driver-latest-dkms 2024-08-06T20:38:19.1406760Z No packages marked for removal. 2024-08-06T20:38:19.1487452Z Dependencies resolved. 2024-08-06T20:38:19.1500784Z Nothing to do. 2024-08-06T20:38:19.1503116Z Complete! 2024-08-06T20:38:19.2782987Z + install_nvidia_driver_common 2024-08-06T20:38:19.2791348Z + echo 'Before installing NVIDIA driver' 2024-08-06T20:38:19.2792684Z Before installing NVIDIA driver 2024-08-06T20:38:19.2793382Z + lspci 2024-08-06T20:38:19.2978681Z 00:00.0 Host bridge: Intel Corporation 440FX - 82441FX PMC [Natoma] (rev 02) 2024-08-06T20:38:19.2979385Z 00:01.0 ISA bridge: Intel Corporation 82371SB PIIX3 ISA [Natoma/Triton II] 2024-08-06T20:38:19.2980060Z 00:01.1 IDE interface: Intel Corporation 82371SB PIIX3 IDE [Natoma/Triton II] 2024-08-06T20:38:19.2980721Z 00:01.3 Bridge: Intel Corporation 82371AB/EB/MB PIIX4 ACPI (rev 01) 2024-08-06T20:38:19.2981287Z 00:02.0 VGA compatible controller: Cirrus Logic GD 5446 2024-08-06T20:38:19.2981879Z 00:03.0 Ethernet controller: Amazon.com, Inc. Elastic Network Adapter (ENA) 2024-08-06T20:38:19.2982611Z 00:1e.0 VGA compatible controller: NVIDIA Corporation GM204GL [Tesla M60] (rev a1) 2024-08-06T20:38:19.2983589Z 00:1f.0 Unassigned class [ff80]: XenSource, Inc. Xen Platform Device (rev 01) 2024-08-06T20:38:19.2984121Z + lsmod 2024-08-06T20:38:19.3022109Z Module Size Used by 2024-08-06T20:38:19.3022573Z veth 36864 0 2024-08-06T20:38:19.3022943Z nvidia_modeset 1351680 0 2024-08-06T20:38:19.3023321Z video 65536 1 nvidia_modeset 2024-08-06T20:38:19.3023745Z wmi 36864 1 video 2024-08-06T20:38:19.3024111Z nvidia_uvm 4706304 0 2024-08-06T20:38:19.3024532Z nvidia 54071296 7 nvidia_uvm,nvidia_modeset 2024-08-06T20:38:19.3024983Z drm 602112 1 nvidia 2024-08-06T20:38:19.3025390Z drm_panel_orientation_quirks 28672 1 drm 2024-08-06T20:38:19.3025882Z backlight 24576 3 video,drm,nvidia_modeset 2024-08-06T20:38:19.3026360Z i2c_core 106496 2 nvidia,drm 2024-08-06T20:38:19.3026768Z xt_conntrack 16384 1 2024-08-06T20:38:19.3027136Z nft_chain_nat 16384 3 2024-08-06T20:38:19.3027504Z xt_MASQUERADE 20480 1 2024-08-06T20:38:19.3027909Z nf_nat 57344 2 nft_chain_nat,xt_MASQUERADE 2024-08-06T20:38:19.3028379Z nf_conntrack_netlink 57344 0 2024-08-06T20:38:19.3028919Z nf_conntrack 184320 4 xt_conntrack,nf_nat,nf_conntrack_netlink,xt_MASQUERADE 2024-08-06T20:38:19.3029492Z nf_defrag_ipv6 24576 1 nf_conntrack 2024-08-06T20:38:19.3029929Z nf_defrag_ipv4 16384 1 nf_conntrack 2024-08-06T20:38:19.3030343Z xfrm_user 57344 1 2024-08-06T20:38:19.3030700Z xfrm_algo 16384 1 xfrm_user 2024-08-06T20:38:19.3031097Z xt_addrtype 16384 2 2024-08-06T20:38:19.3031458Z nft_compat 20480 4 2024-08-06T20:38:19.3031857Z nf_tables 307200 57 nft_compat,nft_chain_nat 2024-08-06T20:38:19.3032416Z nfnetlink 20480 4 nft_compat,nf_conntrack_netlink,nf_tables 2024-08-06T20:38:19.3032933Z br_netfilter 36864 0 2024-08-06T20:38:19.3033300Z bridge 307200 1 br_netfilter 2024-08-06T20:38:19.3033708Z stp 16384 1 bridge 2024-08-06T20:38:19.3034103Z llc 16384 2 bridge,stp 2024-08-06T20:38:19.3034484Z overlay 167936 0 2024-08-06T20:38:19.3034837Z tls 114688 0 2024-08-06T20:38:19.3035188Z ata_piix 45056 0 2024-08-06T20:38:19.3035536Z libata 401408 1 ata_piix 2024-08-06T20:38:19.3035924Z nls_ascii 16384 1 2024-08-06T20:38:19.3036276Z ena 167936 0 2024-08-06T20:38:19.3036612Z scsi_mod 290816 1 libata 2024-08-06T20:38:19.3036995Z nls_cp437 20480 1 2024-08-06T20:38:19.3037350Z ptp 36864 1 ena 2024-08-06T20:38:19.3037742Z scsi_common 16384 2 scsi_mod,libata 2024-08-06T20:38:19.3038165Z ghash_clmulni_intel 16384 0 2024-08-06T20:38:19.3038542Z pps_core 24576 1 ptp 2024-08-06T20:38:19.3038909Z vfat 24576 1 2024-08-06T20:38:19.3039265Z aesni_intel 393216 0 2024-08-06T20:38:19.3039621Z fat 86016 1 vfat 2024-08-06T20:38:19.3039976Z i8042 45056 0 2024-08-06T20:38:19.3040341Z serio 28672 3 i8042 2024-08-06T20:38:19.3040745Z crypto_simd 16384 1 aesni_intel 2024-08-06T20:38:19.3041206Z cryptd 28672 2 crypto_simd,ghash_clmulni_intel 2024-08-06T20:38:19.3041681Z button 24576 0 2024-08-06T20:38:19.3042041Z sunrpc 692224 1 2024-08-06T20:38:19.3042383Z sch_fq_codel 20480 9 2024-08-06T20:38:19.3042741Z dm_mod 188416 0 2024-08-06T20:38:19.3043093Z fuse 163840 1 2024-08-06T20:38:19.3043432Z dax 45056 1 dm_mod 2024-08-06T20:38:19.3043818Z configfs 57344 1 2024-08-06T20:38:19.3044174Z loop 36864 0 2024-08-06T20:38:19.3044513Z dmi_sysfs 20480 0 2024-08-06T20:38:19.3045018Z crc32_pclmul 16384 0 2024-08-06T20:38:19.3045364Z crc32c_intel 24576 0 2024-08-06T20:38:19.3045745Z + modinfo nvidia 2024-08-06T20:38:19.3046350Z filename: /lib/modules/6.1.94-99.176.amzn2023.x86_64/kernel/drivers/video/nvidia.ko 2024-08-06T20:38:19.3046960Z alias: char-major-195-* 2024-08-06T20:38:19.3047322Z version: 550.54.15 2024-08-06T20:38:19.3047670Z supported: external 2024-08-06T20:38:19.3048016Z license: NVIDIA 2024-08-06T20:38:19.3048371Z firmware: nvidia/550.54.15/gsp_tu10x.bin 2024-08-06T20:38:19.3048838Z firmware: nvidia/550.54.15/gsp_ga10x.bin 2024-08-06T20:38:19.3049273Z srcversion: 833721318DA517F0C2FEC97 2024-08-06T20:38:19.3049694Z alias: pci:v000010DEd*sv*sd*bc06sc80i00* 2024-08-06T20:38:19.3050160Z alias: pci:v000010DEd*sv*sd*bc03sc02i00* 2024-08-06T20:38:19.3050620Z alias: pci:v000010DEd*sv*sd*bc03sc00i00* 2024-08-06T20:38:19.3051041Z depends: i2c-core,drm 2024-08-06T20:38:19.3051397Z retpoline: Y 2024-08-06T20:38:19.3051707Z name: nvidia 2024-08-06T20:38:19.3052176Z vermagic: 6.1.94-99.176.amzn2023.x86_64 SMP preempt mod_unload modversions 2024-08-06T20:38:19.3052804Z parm: NvSwitchRegDwords:NvSwitch regkey (charp) 2024-08-06T20:38:19.3053406Z parm: NvSwitchBlacklist:NvSwitchBlacklist=uuid[,uuid...] (charp) 2024-08-06T20:38:19.3053954Z parm: NVreg_ResmanDebugLevel:int 2024-08-06T20:38:19.3054380Z parm: NVreg_RmLogonRC:int 2024-08-06T20:38:19.3054795Z parm: NVreg_ModifyDeviceFiles:int 2024-08-06T20:38:19.3055212Z parm: NVreg_DeviceFileUID:int 2024-08-06T20:38:19.3055626Z parm: NVreg_DeviceFileGID:int 2024-08-06T20:38:19.3056044Z parm: NVreg_DeviceFileMode:int 2024-08-06T20:38:19.3056530Z parm: NVreg_InitializeSystemMemoryAllocations:int 2024-08-06T20:38:19.3057053Z parm: NVreg_UsePageAttributeTable:int 2024-08-06T20:38:19.3057518Z parm: NVreg_EnablePCIeGen3:int 2024-08-06T20:38:19.3057917Z parm: NVreg_EnableMSI:int 2024-08-06T20:38:19.3058315Z parm: NVreg_TCEBypassMode:int 2024-08-06T20:38:19.3058752Z parm: NVreg_EnableStreamMemOPs:int 2024-08-06T20:38:19.3059236Z parm: NVreg_RestrictProfilingToAdminUsers:int 2024-08-06T20:38:19.3059774Z parm: NVreg_PreserveVideoMemoryAllocations:int 2024-08-06T20:38:19.3060290Z parm: NVreg_EnableS0ixPowerManagement:int 2024-08-06T20:38:19.3060826Z parm: NVreg_S0ixPowerManagementVideoMemoryThreshold:int 2024-08-06T20:38:19.3061372Z parm: NVreg_DynamicPowerManagement:int 2024-08-06T20:38:19.3061939Z parm: NVreg_DynamicPowerManagementVideoMemoryThreshold:int 2024-08-06T20:38:19.3062484Z parm: NVreg_EnableGpuFirmware:int 2024-08-06T20:38:19.3062949Z parm: NVreg_EnableGpuFirmwareLogs:int 2024-08-06T20:38:19.3063461Z parm: NVreg_OpenRmEnableUnsupportedGpus:int 2024-08-06T20:38:19.3063960Z parm: NVreg_EnableUserNUMAManagement:int 2024-08-06T20:38:19.3064429Z parm: NVreg_MemoryPoolSize:int 2024-08-06T20:38:19.3064875Z parm: NVreg_KMallocHeapMaxSize:int 2024-08-06T20:38:19.3065321Z parm: NVreg_VMallocHeapMaxSize:int 2024-08-06T20:38:19.3065772Z parm: NVreg_IgnoreMMIOCheck:int 2024-08-06T20:38:19.3066203Z parm: NVreg_NvLinkDisable:int 2024-08-06T20:38:19.3066657Z parm: NVreg_EnablePCIERelaxedOrderingMode:int 2024-08-06T20:38:19.3067152Z parm: NVreg_RegisterPCIDriver:int 2024-08-06T20:38:19.3067605Z parm: NVreg_EnableResizableBar:int 2024-08-06T20:38:19.3068049Z parm: NVreg_EnableDbgBreakpoint:int 2024-08-06T20:38:19.3068522Z parm: NVreg_EnableNonblockingOpen:int 2024-08-06T20:38:19.3068984Z parm: NVreg_RegistryDwords:charp 2024-08-06T20:38:19.3069435Z parm: NVreg_RegistryDwordsPerDevice:charp 2024-08-06T20:38:19.3069887Z parm: NVreg_RmMsg:charp 2024-08-06T20:38:19.3070434Z parm: NVreg_GpuBlacklist:charp 2024-08-06T20:38:19.3070875Z parm: NVreg_TemporaryFilePath:charp 2024-08-06T20:38:19.3071324Z parm: NVreg_ExcludedGpus:charp 2024-08-06T20:38:19.3071844Z parm: NVreg_DmaRemapPeerMmio:int 2024-08-06T20:38:19.3072293Z parm: NVreg_RmNvlinkBandwidth:charp 2024-08-06T20:38:19.3072765Z parm: NVreg_ImexChannelCount:int 2024-08-06T20:38:19.3073197Z parm: rm_firmware_active:charp 2024-08-06T20:38:19.3073585Z + HAS_NVIDIA_DRIVER=0 2024-08-06T20:38:19.3073929Z ++ command -v nvidia-smi 2024-08-06T20:38:19.3074297Z + '[' -x /usr/bin/nvidia-smi ']' 2024-08-06T20:38:19.3074643Z + set +e 2024-08-06T20:38:19.3075069Z ++ nvidia-smi --query-gpu=driver_version --format=csv,noheader --id=0 2024-08-06T20:38:19.3316774Z + INSTALLED_DRIVER_VERSION=550.54.15 2024-08-06T20:38:19.3317180Z + NVIDIA_SMI_STATUS=0 2024-08-06T20:38:19.3317514Z + '[' 0 -ne 0 ']' 2024-08-06T20:38:19.3317899Z + '[' 550.54.15 '!=' 550.54.15 ']' 2024-08-06T20:38:19.3318604Z + HAS_NVIDIA_DRIVER=1 2024-08-06T20:38:19.3319675Z + echo 'NVIDIA driver (550.54.15) has already been installed. Skipping NVIDIA driver installation' 2024-08-06T20:38:19.3320849Z + set -e 2024-08-06T20:38:19.3321341Z + '[' 1 -eq 0 ']' 2024-08-06T20:38:19.3322260Z NVIDIA driver (550.54.15) has already been installed. Skipping NVIDIA driver installation 2024-08-06T20:38:19.3322888Z + post_install_nvidia_driver_common 2024-08-06T20:38:19.3324823Z + sudo modprobe nvidia 2024-08-06T20:38:19.4911484Z + echo 'After installing NVIDIA driver' 2024-08-06T20:38:19.4912157Z + lspci 2024-08-06T20:38:19.4912472Z After installing NVIDIA driver 2024-08-06T20:38:19.5101320Z 00:00.0 Host bridge: Intel Corporation 440FX - 82441FX PMC [Natoma] (rev 02) 2024-08-06T20:38:19.5102129Z 00:01.0 ISA bridge: Intel Corporation 82371SB PIIX3 ISA [Natoma/Triton II] 2024-08-06T20:38:19.5102998Z 00:01.1 IDE interface: Intel Corporation 82371SB PIIX3 IDE [Natoma/Triton II] 2024-08-06T20:38:19.5103675Z 00:01.3 Bridge: Intel Corporation 82371AB/EB/MB PIIX4 ACPI (rev 01) 2024-08-06T20:38:19.5104684Z 00:02.0 VGA compatible controller: Cirrus Logic GD 5446 2024-08-06T20:38:19.5105317Z 00:03.0 Ethernet controller: Amazon.com, Inc. Elastic Network Adapter (ENA) 2024-08-06T20:38:19.5106729Z 00:1e.0 VGA compatible controller: NVIDIA Corporation GM204GL [Tesla M60] (rev a1) 2024-08-06T20:38:19.5107734Z 00:1f.0 Unassigned class [ff80]: XenSource, Inc. Xen Platform Device (rev 01) 2024-08-06T20:38:19.5108282Z + lsmod 2024-08-06T20:38:19.5131902Z Module Size Used by 2024-08-06T20:38:19.5132703Z veth 36864 0 2024-08-06T20:38:19.5133490Z nvidia_modeset 1351680 0 2024-08-06T20:38:19.5134205Z video 65536 1 nvidia_modeset 2024-08-06T20:38:19.5134612Z wmi 36864 1 video 2024-08-06T20:38:19.5134991Z nvidia_uvm 4706304 0 2024-08-06T20:38:19.5135406Z nvidia 54071296 7 nvidia_uvm,nvidia_modeset 2024-08-06T20:38:19.5135858Z drm 602112 1 nvidia 2024-08-06T20:38:19.5136280Z drm_panel_orientation_quirks 28672 1 drm 2024-08-06T20:38:19.5136776Z backlight 24576 3 video,drm,nvidia_modeset 2024-08-06T20:38:19.5137241Z i2c_core 106496 2 nvidia,drm 2024-08-06T20:38:19.5137648Z xt_conntrack 16384 1 2024-08-06T20:38:19.5138014Z nft_chain_nat 16384 3 2024-08-06T20:38:19.5138363Z xt_MASQUERADE 20480 1 2024-08-06T20:38:19.5138779Z nf_nat 57344 2 nft_chain_nat,xt_MASQUERADE 2024-08-06T20:38:19.5139237Z nf_conntrack_netlink 57344 0 2024-08-06T20:38:19.5139762Z nf_conntrack 184320 4 xt_conntrack,nf_nat,nf_conntrack_netlink,xt_MASQUERADE 2024-08-06T20:38:19.5140362Z nf_defrag_ipv6 24576 1 nf_conntrack 2024-08-06T20:38:19.5140802Z nf_defrag_ipv4 16384 1 nf_conntrack 2024-08-06T20:38:19.5141197Z xfrm_user 57344 1 2024-08-06T20:38:19.5141570Z xfrm_algo 16384 1 xfrm_user 2024-08-06T20:38:19.5142221Z xt_addrtype 16384 2 2024-08-06T20:38:19.5142568Z nft_compat 20480 4 2024-08-06T20:38:19.5142990Z nf_tables 307200 57 nft_compat,nft_chain_nat 2024-08-06T20:38:19.5143740Z nfnetlink 20480 4 nft_compat,nf_conntrack_netlink,nf_tables 2024-08-06T20:38:19.5144258Z br_netfilter 36864 0 2024-08-06T20:38:19.5144647Z bridge 307200 1 br_netfilter 2024-08-06T20:38:19.5145061Z stp 16384 1 bridge 2024-08-06T20:38:19.5145439Z llc 16384 2 bridge,stp 2024-08-06T20:38:19.5145837Z overlay 167936 0 2024-08-06T20:38:19.5146188Z tls 114688 0 2024-08-06T20:38:19.5146523Z ata_piix 45056 0 2024-08-06T20:38:19.5146888Z libata 401408 1 ata_piix 2024-08-06T20:38:19.5147459Z nls_ascii 16384 1 2024-08-06T20:38:19.5148197Z ena 167936 0 2024-08-06T20:38:19.5148926Z scsi_mod 290816 1 libata 2024-08-06T20:38:19.5149343Z nls_cp437 20480 1 2024-08-06T20:38:19.5149685Z ptp 36864 1 ena 2024-08-06T20:38:19.5150096Z scsi_common 16384 2 scsi_mod,libata 2024-08-06T20:38:19.5150526Z ghash_clmulni_intel 16384 0 2024-08-06T20:38:19.5150890Z pps_core 24576 1 ptp 2024-08-06T20:38:19.5151264Z vfat 24576 1 2024-08-06T20:38:19.5151904Z aesni_intel 393216 0 2024-08-06T20:38:19.5152679Z fat 86016 1 vfat 2024-08-06T20:38:19.5153143Z i8042 45056 0 2024-08-06T20:38:19.5153482Z serio 28672 3 i8042 2024-08-06T20:38:19.5153883Z crypto_simd 16384 1 aesni_intel 2024-08-06T20:38:19.5154358Z cryptd 28672 2 crypto_simd,ghash_clmulni_intel 2024-08-06T20:38:19.5154806Z button 24576 0 2024-08-06T20:38:19.5155168Z sunrpc 692224 1 2024-08-06T20:38:19.5155524Z sch_fq_codel 20480 9 2024-08-06T20:38:19.5155865Z dm_mod 188416 0 2024-08-06T20:38:19.5156229Z fuse 163840 1 2024-08-06T20:38:19.5156582Z dax 45056 1 dm_mod 2024-08-06T20:38:19.5156946Z configfs 57344 1 2024-08-06T20:38:19.5157314Z loop 36864 0 2024-08-06T20:38:19.5157674Z dmi_sysfs 20480 0 2024-08-06T20:38:19.5158015Z crc32_pclmul 16384 0 2024-08-06T20:38:19.5158378Z crc32c_intel 24576 0 2024-08-06T20:38:19.5158762Z + modinfo nvidia 2024-08-06T20:38:19.5159255Z filename: /lib/modules/6.1.94-99.176.amzn2023.x86_64/kernel/drivers/video/nvidia.ko 2024-08-06T20:38:19.5159856Z alias: char-major-195-* 2024-08-06T20:38:19.5160235Z version: 550.54.15 2024-08-06T20:38:19.5160563Z supported: external 2024-08-06T20:38:19.5160905Z license: NVIDIA 2024-08-06T20:38:19.5161277Z firmware: nvidia/550.54.15/gsp_tu10x.bin 2024-08-06T20:38:19.5161729Z firmware: nvidia/550.54.15/gsp_ga10x.bin 2024-08-06T20:38:19.5162182Z srcversion: 833721318DA517F0C2FEC97 2024-08-06T20:38:19.5162622Z alias: pci:v000010DEd*sv*sd*bc06sc80i00* 2024-08-06T20:38:19.5163074Z alias: pci:v000010DEd*sv*sd*bc03sc02i00* 2024-08-06T20:38:19.5163545Z alias: pci:v000010DEd*sv*sd*bc03sc00i00* 2024-08-06T20:38:19.5163978Z depends: i2c-core,drm 2024-08-06T20:38:19.5164324Z retpoline: Y 2024-08-06T20:38:19.5164636Z name: nvidia 2024-08-06T20:38:19.5165114Z vermagic: 6.1.94-99.176.amzn2023.x86_64 SMP preempt mod_unload modversions 2024-08-06T20:38:19.5165720Z parm: NvSwitchRegDwords:NvSwitch regkey (charp) 2024-08-06T20:38:19.5166326Z parm: NvSwitchBlacklist:NvSwitchBlacklist=uuid[,uuid...] (charp) 2024-08-06T20:38:19.5166896Z parm: NVreg_ResmanDebugLevel:int 2024-08-06T20:38:19.5167304Z parm: NVreg_RmLogonRC:int 2024-08-06T20:38:19.5167720Z parm: NVreg_ModifyDeviceFiles:int 2024-08-06T20:38:19.5168149Z parm: NVreg_DeviceFileUID:int 2024-08-06T20:38:19.5168710Z parm: NVreg_DeviceFileGID:int 2024-08-06T20:38:19.5169128Z parm: NVreg_DeviceFileMode:int 2024-08-06T20:38:19.5169613Z parm: NVreg_InitializeSystemMemoryAllocations:int 2024-08-06T20:38:19.5170204Z parm: NVreg_UsePageAttributeTable:int 2024-08-06T20:38:19.5170724Z parm: NVreg_EnablePCIeGen3:int 2024-08-06T20:38:19.5171137Z parm: NVreg_EnableMSI:int 2024-08-06T20:38:19.5171521Z parm: NVreg_TCEBypassMode:int 2024-08-06T20:38:19.5171956Z parm: NVreg_EnableStreamMemOPs:int 2024-08-06T20:38:19.5172446Z parm: NVreg_RestrictProfilingToAdminUsers:int 2024-08-06T20:38:19.5172964Z parm: NVreg_PreserveVideoMemoryAllocations:int 2024-08-06T20:38:19.5173481Z parm: NVreg_EnableS0ixPowerManagement:int 2024-08-06T20:38:19.5174035Z parm: NVreg_S0ixPowerManagementVideoMemoryThreshold:int 2024-08-06T20:38:19.5174565Z parm: NVreg_DynamicPowerManagement:int 2024-08-06T20:38:19.5175137Z parm: NVreg_DynamicPowerManagementVideoMemoryThreshold:int 2024-08-06T20:38:19.5175692Z parm: NVreg_EnableGpuFirmware:int 2024-08-06T20:38:19.5176138Z parm: NVreg_EnableGpuFirmwareLogs:int 2024-08-06T20:38:19.5176647Z parm: NVreg_OpenRmEnableUnsupportedGpus:int 2024-08-06T20:38:19.5177154Z parm: NVreg_EnableUserNUMAManagement:int 2024-08-06T20:38:19.5177604Z parm: NVreg_MemoryPoolSize:int 2024-08-06T20:38:19.5178047Z parm: NVreg_KMallocHeapMaxSize:int 2024-08-06T20:38:19.5178499Z parm: NVreg_VMallocHeapMaxSize:int 2024-08-06T20:38:19.5178929Z parm: NVreg_IgnoreMMIOCheck:int 2024-08-06T20:38:19.5179360Z parm: NVreg_NvLinkDisable:int 2024-08-06T20:38:19.5179834Z parm: NVreg_EnablePCIERelaxedOrderingMode:int 2024-08-06T20:38:19.5180310Z parm: NVreg_RegisterPCIDriver:int 2024-08-06T20:38:19.5180763Z parm: NVreg_EnableResizableBar:int 2024-08-06T20:38:19.5181225Z parm: NVreg_EnableDbgBreakpoint:int 2024-08-06T20:38:19.5181689Z parm: NVreg_EnableNonblockingOpen:int 2024-08-06T20:38:19.5182156Z parm: NVreg_RegistryDwords:charp 2024-08-06T20:38:19.5182623Z parm: NVreg_RegistryDwordsPerDevice:charp 2024-08-06T20:38:19.5183066Z parm: NVreg_RmMsg:charp 2024-08-06T20:38:19.5183477Z parm: NVreg_GpuBlacklist:charp 2024-08-06T20:38:19.5183925Z parm: NVreg_TemporaryFilePath:charp 2024-08-06T20:38:19.5184359Z parm: NVreg_ExcludedGpus:charp 2024-08-06T20:38:19.5184801Z parm: NVreg_DmaRemapPeerMmio:int 2024-08-06T20:38:19.5185254Z parm: NVreg_RmNvlinkBandwidth:charp 2024-08-06T20:38:19.5185694Z parm: NVreg_ImexChannelCount:int 2024-08-06T20:38:19.5186131Z parm: rm_firmware_active:charp 2024-08-06T20:38:19.5186598Z + set +e 2024-08-06T20:38:19.5187056Z + nvidia-smi 2024-08-06T20:38:19.5359082Z Tue Aug 6 20:38:19 2024 2024-08-06T20:38:19.5359599Z +-----------------------------------------------------------------------------------------+ 2024-08-06T20:38:19.5360287Z | NVIDIA-SMI 550.54.15 Driver Version: 550.54.15 CUDA Version: 12.4 | 2024-08-06T20:38:19.5360958Z |-----------------------------------------+------------------------+----------------------+ 2024-08-06T20:38:19.5361618Z | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | 2024-08-06T20:38:19.5362308Z | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | 2024-08-06T20:38:19.5362894Z | | | MIG M. | 2024-08-06T20:38:19.5363353Z |=========================================+========================+======================| 2024-08-06T20:38:19.5446954Z | 0 Tesla M60 On | 00000000:00:1E.0 Off | 0 | 2024-08-06T20:38:19.5447598Z | N/A 32C P8 15W / 150W | 0MiB / 7680MiB | 0% Default | 2024-08-06T20:38:19.5448428Z | | | N/A | 2024-08-06T20:38:19.5448959Z +-----------------------------------------+------------------------+----------------------+ 2024-08-06T20:38:19.5449653Z 2024-08-06T20:38:19.5450239Z +-----------------------------------------------------------------------------------------+ 2024-08-06T20:38:19.5450807Z | Processes: | 2024-08-06T20:38:19.5451417Z | GPU GI CI PID Type Process name GPU Memory | 2024-08-06T20:38:19.5451993Z | ID ID Usage | 2024-08-06T20:38:19.5452465Z |=========================================================================================| 2024-08-06T20:38:19.5453027Z | No running processes found | 2024-08-06T20:38:19.5453674Z +-----------------------------------------------------------------------------------------+ 2024-08-06T20:38:19.6041797Z + nvidia-smi --query-gpu=gpu_name --format=csv,noheader --id=0 2024-08-06T20:38:19.6229211Z Tesla M60 2024-08-06T20:38:19.6291206Z + NVIDIA_SMI_STATUS=0 2024-08-06T20:38:19.6291576Z + '[' 0 -eq 0 ']' 2024-08-06T20:38:19.6291901Z + echo 'INFO: Ignoring allowed status 0' 2024-08-06T20:38:19.6292309Z + set -e 2024-08-06T20:38:19.6292616Z INFO: Ignoring allowed status 0 2024-08-06T20:38:19.6301381Z == Installing nvidia container toolkit for amzn2023 == 2024-08-06T20:38:19.6306215Z + sudo yum install -y yum-utils 2024-08-06T20:38:20.1598125Z Last metadata expiration check: 2:54:20 ago on Tue Aug 6 17:44:00 2024. 2024-08-06T20:38:20.1894159Z Package dnf-utils-4.3.0-13.amzn2023.0.4.noarch is already installed. 2024-08-06T20:38:20.2317101Z Dependencies resolved. 2024-08-06T20:38:20.2502791Z Nothing to do. 2024-08-06T20:38:20.2503665Z Complete! 2024-08-06T20:38:20.3737107Z + [[ amzn2023 == \a\m\z\n\2\0\2\3 ]] 2024-08-06T20:38:20.3738509Z + YUM_REPO_URL=https://nvidia.github.io/libnvidia-container/stable/rpm/nvidia-container-toolkit.repo 2024-08-06T20:38:20.3740823Z + sudo yum-config-manager --add-repo https://nvidia.github.io/libnvidia-container/stable/rpm/nvidia-container-toolkit.repo 2024-08-06T20:38:20.7128979Z Adding repo from: https://nvidia.github.io/libnvidia-container/stable/rpm/nvidia-container-toolkit.repo 2024-08-06T20:38:20.9013661Z + sudo yum install -y nvidia-docker2 2024-08-06T20:38:21.4998298Z nvidia-container-toolkit 8.6 kB/s | 833 B 00:00 2024-08-06T20:38:21.5301176Z Package nvidia-docker2-2.14.0-1.noarch is already installed. 2024-08-06T20:38:21.5714523Z Dependencies resolved. 2024-08-06T20:38:21.5899235Z Nothing to do. 2024-08-06T20:38:21.5899811Z Complete! 2024-08-06T20:38:21.7423657Z + sudo systemctl restart docker 2024-08-06T20:38:45.4007049Z nvidia-persistenced failed to initialize. Check syslog for more details. 2024-08-06T20:38:45.4237743Z Tue Aug 6 20:38:45 2024 2024-08-06T20:38:45.4238279Z +-----------------------------------------------------------------------------------------+ 2024-08-06T20:38:45.4239009Z | NVIDIA-SMI 550.54.15 Driver Version: 550.54.15 CUDA Version: 12.4 | 2024-08-06T20:38:45.4239673Z |-----------------------------------------+------------------------+----------------------+ 2024-08-06T20:38:45.4240331Z | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | 2024-08-06T20:38:45.4241048Z | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | 2024-08-06T20:38:45.4241632Z | | | MIG M. | 2024-08-06T20:38:45.4242081Z |=========================================+========================+======================| 2024-08-06T20:38:45.4325009Z | 0 Tesla M60 On | 00000000:00:1E.0 Off | 0 | 2024-08-06T20:38:45.4326551Z | N/A 32C P8 15W / 150W | 0MiB / 7680MiB | 0% Default | 2024-08-06T20:38:45.4327820Z | | | N/A | 2024-08-06T20:38:45.4328782Z +-----------------------------------------+------------------------+----------------------+ 2024-08-06T20:38:45.4329710Z 2024-08-06T20:38:45.4330587Z +-----------------------------------------------------------------------------------------+ 2024-08-06T20:38:45.4331624Z | Processes: | 2024-08-06T20:38:45.4332742Z | GPU GI CI PID Type Process name GPU Memory | 2024-08-06T20:38:45.4333689Z | ID ID Usage | 2024-08-06T20:38:45.4334496Z |=========================================================================================| 2024-08-06T20:38:45.4335493Z | No running processes found | 2024-08-06T20:38:45.4336694Z +-----------------------------------------------------------------------------------------+ 2024-08-06T20:38:45.6987649Z Command completed after 1 attempt(s). 2024-08-06T20:38:45.7072736Z ##[group]Run python3 -m pip install psutil==5.9.1 nvidia-ml-py==11.525.84 2024-08-06T20:38:45.7073580Z python3 -m pip install psutil==5.9.1 nvidia-ml-py==11.525.84 2024-08-06T20:38:45.7074224Z python3 -m tools.stats.monitor > usage_log.txt 2>&1 & 2024-08-06T20:38:45.7074826Z echo "monitor-script-pid=${!}" >> "${GITHUB_OUTPUT}" 2024-08-06T20:38:45.7084725Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:38:45.7085638Z env: 2024-08-06T20:38:45.7086143Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:38:45.7086936Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:38:45.7087807Z ##[endgroup] 2024-08-06T20:38:46.0411849Z Defaulting to user installation because normal site-packages is not writeable 2024-08-06T20:38:46.0622229Z Requirement already satisfied: psutil==5.9.1 in /home/ec2-user/.local/lib/python3.9/site-packages (5.9.1) 2024-08-06T20:38:46.0630626Z Requirement already satisfied: nvidia-ml-py==11.525.84 in /home/ec2-user/.local/lib/python3.9/site-packages (11.525.84) 2024-08-06T20:38:46.2209288Z Prepare all required actions 2024-08-06T20:38:46.2210022Z Getting action download info 2024-08-06T20:38:46.3290883Z Download action repository 'seemethere/download-artifact-s3@v4' (SHA:1da556a7aa0a088e3153970611f6c432d58e80e6) 2024-08-06T20:38:48.4691617Z Download action repository 'actions/download-artifact@v3' (SHA:9bc31d5ccc31df68ecc42ccf4149144866c47d8a) 2024-08-06T20:38:48.6709525Z ##[group]Run ./.github/actions/download-build-artifacts 2024-08-06T20:38:48.6709992Z with: 2024-08-06T20:38:48.6710340Z name: linux-focal-cuda12.1-py3.10-gcc9 2024-08-06T20:38:48.6710744Z s3-bucket: gha-artifacts 2024-08-06T20:38:48.6711086Z env: 2024-08-06T20:38:48.6711367Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:38:48.6711780Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:38:48.6712240Z ##[endgroup] 2024-08-06T20:38:48.6761320Z ##[group]Run seemethere/download-artifact-s3@v4 2024-08-06T20:38:48.6761768Z with: 2024-08-06T20:38:48.6762136Z name: linux-focal-cuda12.1-py3.10-gcc9 2024-08-06T20:38:48.6762565Z s3-bucket: gha-artifacts 2024-08-06T20:38:48.6762902Z region: us-east-1 2024-08-06T20:38:48.6763203Z env: 2024-08-06T20:38:48.6763467Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:38:48.6763899Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:38:48.6764363Z ##[endgroup] 2024-08-06T20:38:49.3107067Z (node:205341) NOTE: We are formalizing our plans to enter AWS SDK for JavaScript (v2) into maintenance mode in 2023. 2024-08-06T20:38:49.3107683Z 2024-08-06T20:38:49.3108320Z Please migrate your code to use AWS SDK for JavaScript (v3). 2024-08-06T20:38:49.3108984Z For more information, check the migration guide at https://a.co/7PzMCcy 2024-08-06T20:38:49.3109686Z (Use `node --trace-warnings ...` to show where the warning was created) 2024-08-06T20:38:49.4089214Z Found 1 objects with prefix pytorch/pytorch/10273124344/linux-focal-cuda12.1-py3.10-gcc9/ 2024-08-06T20:38:49.4090117Z Starting download (1/1): /home/ec2-user/actions-runner/_work/pytorch/pytorch/artifacts.zip 2024-08-06T20:39:01.9553160Z Finished download (1/1): /home/ec2-user/actions-runner/_work/pytorch/pytorch/artifacts.zip 2024-08-06T20:39:01.9564988Z Artifact download has finished successfully 2024-08-06T20:39:01.9821031Z ##[group]Run unzip -o artifacts.zip 2024-08-06T20:39:01.9821480Z unzip -o artifacts.zip 2024-08-06T20:39:01.9828613Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:39:01.9829103Z env: 2024-08-06T20:39:01.9829396Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:01.9829834Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:01.9830301Z ##[endgroup] 2024-08-06T20:39:01.9875269Z Archive: artifacts.zip 2024-08-06T20:39:01.9877029Z creating: dist/ 2024-08-06T20:39:04.3838990Z inflating: dist/torch-2.5.0a0+gitb9d86fa-cp310-cp310-linux_x86_64.whl 2024-08-06T20:39:04.3839633Z creating: build/custom_test_artifacts/ 2024-08-06T20:39:04.3840140Z creating: build/custom_test_artifacts/custom-op-build/ 2024-08-06T20:39:04.3840773Z creating: build/custom_test_artifacts/custom-op-build/CMakeFiles/ 2024-08-06T20:39:04.3841517Z creating: build/custom_test_artifacts/custom-op-build/CMakeFiles/pkgRedirects/ 2024-08-06T20:39:04.3849727Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/CMakeConfigureLog.yaml 2024-08-06T20:39:04.3850568Z creating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/ 2024-08-06T20:39:04.3851613Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CMakeSystem.cmake 2024-08-06T20:39:04.3852545Z creating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdC/ 2024-08-06T20:39:04.3853473Z creating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdC/tmp/ 2024-08-06T20:39:04.3856446Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdC/CMakeCCompilerId.c 2024-08-06T20:39:04.3858411Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdC/a.out 2024-08-06T20:39:04.3859360Z creating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCXX/ 2024-08-06T20:39:04.3860244Z creating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCXX/tmp/ 2024-08-06T20:39:04.3862946Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCXX/CMakeCXXCompilerId.cpp 2024-08-06T20:39:04.3864567Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCXX/a.out 2024-08-06T20:39:04.3866963Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CMakeDetermineCompilerABI_C.bin 2024-08-06T20:39:04.3867995Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CMakeCCompiler.cmake 2024-08-06T20:39:04.3869640Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CMakeDetermineCompilerABI_CXX.bin 2024-08-06T20:39:04.3871123Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CMakeCXXCompiler.cmake 2024-08-06T20:39:04.3872089Z creating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/ 2024-08-06T20:39:04.3873067Z creating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/ 2024-08-06T20:39:04.3920526Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cpp4.ii 2024-08-06T20:39:04.3966951Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cudafe1.cpp 2024-08-06T20:39:04.3968630Z extracting: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.module_id 2024-08-06T20:39:04.4022286Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cpp1.ii 2024-08-06T20:39:04.4023664Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cudafe1.c 2024-08-06T20:39:04.4025077Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cudafe1.gpu 2024-08-06T20:39:04.4026530Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cudafe1.stub.c 2024-08-06T20:39:04.4027941Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.ptx 2024-08-06T20:39:04.4029312Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.sm_52.cubin 2024-08-06T20:39:04.4030701Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.fatbin 2024-08-06T20:39:04.4032072Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.fatbin.c 2024-08-06T20:39:04.4033431Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.o 2024-08-06T20:39:04.4034688Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.sm_52.cubin 2024-08-06T20:39:04.4035927Z extracting: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.reg.c 2024-08-06T20:39:04.4037135Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.fatbin 2024-08-06T20:39:04.4038365Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.fatbin.c 2024-08-06T20:39:04.4039516Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.o 2024-08-06T20:39:04.4040717Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/CMakeCUDACompilerId.cu 2024-08-06T20:39:04.4110979Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CompilerIdCUDA/a.out 2024-08-06T20:39:04.4183370Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CMakeDetermineCompilerABI_CUDA.bin 2024-08-06T20:39:04.4184436Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/3.26.4/CMakeCUDACompiler.cmake 2024-08-06T20:39:04.4185409Z creating: build/custom_test_artifacts/custom-op-build/CMakeFiles/CMakeScratch/ 2024-08-06T20:39:04.4186246Z creating: build/custom_test_artifacts/custom-op-build/CMakeFiles/CMakeTmp/ 2024-08-06T20:39:04.4187125Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/cmake.check_cache 2024-08-06T20:39:04.4188027Z creating: build/custom_test_artifacts/custom-op-build/CMakeFiles/custom_ops.dir/ 2024-08-06T20:39:04.4189026Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/custom_ops.dir/compiler_depend.ts 2024-08-06T20:39:04.4190146Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/custom_ops.dir/compiler_depend.make 2024-08-06T20:39:04.4191256Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/custom_ops.dir/depend.make 2024-08-06T20:39:04.4192272Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/custom_ops.dir/link.txt 2024-08-06T20:39:04.4193332Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/custom_ops.dir/cmake_clean.cmake 2024-08-06T20:39:04.4194390Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/custom_ops.dir/build.make 2024-08-06T20:39:04.4195624Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/custom_ops.dir/DependInfo.cmake 2024-08-06T20:39:04.4196710Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/custom_ops.dir/flags.make 2024-08-06T20:39:04.4197740Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/custom_ops.dir/progress.make 2024-08-06T20:39:04.4219149Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/custom_ops.dir/op.cpp.o.d 2024-08-06T20:39:04.4390293Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/custom_ops.dir/op.cpp.o 2024-08-06T20:39:04.4391191Z creating: build/custom_test_artifacts/custom-op-build/CMakeFiles/test_custom_ops.dir/ 2024-08-06T20:39:04.4392144Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/test_custom_ops.dir/compiler_depend.ts 2024-08-06T20:39:04.4393235Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/test_custom_ops.dir/compiler_depend.make 2024-08-06T20:39:04.4394290Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/test_custom_ops.dir/depend.make 2024-08-06T20:39:04.4395269Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/test_custom_ops.dir/link.txt 2024-08-06T20:39:04.4396267Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/test_custom_ops.dir/cmake_clean.cmake 2024-08-06T20:39:04.4397304Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/test_custom_ops.dir/build.make 2024-08-06T20:39:04.4398323Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/test_custom_ops.dir/DependInfo.cmake 2024-08-06T20:39:04.4399324Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/test_custom_ops.dir/flags.make 2024-08-06T20:39:04.4400329Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/test_custom_ops.dir/progress.make 2024-08-06T20:39:04.4423830Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/test_custom_ops.dir/test_custom_ops.cpp.o.d 2024-08-06T20:39:04.4523629Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/test_custom_ops.dir/test_custom_ops.cpp.o 2024-08-06T20:39:04.4524708Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/CMakeDirectoryInformation.cmake 2024-08-06T20:39:04.4525779Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/TargetDirectories.txt 2024-08-06T20:39:04.4526914Z extracting: build/custom_test_artifacts/custom-op-build/CMakeFiles/progress.marks 2024-08-06T20:39:04.4527843Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/Makefile2 2024-08-06T20:39:04.4528714Z inflating: build/custom_test_artifacts/custom-op-build/CMakeFiles/Makefile.cmake 2024-08-06T20:39:04.4529607Z inflating: build/custom_test_artifacts/custom-op-build/detect_cuda_version.cc 2024-08-06T20:39:04.4532540Z inflating: build/custom_test_artifacts/custom-op-build/CMakeCache.txt 2024-08-06T20:39:04.4533423Z inflating: build/custom_test_artifacts/custom-op-build/Makefile 2024-08-06T20:39:04.4534306Z inflating: build/custom_test_artifacts/custom-op-build/cmake_install.cmake 2024-08-06T20:39:04.4678505Z inflating: build/custom_test_artifacts/custom-op-build/libcustom_ops.so 2024-08-06T20:39:04.4754315Z inflating: build/custom_test_artifacts/custom-op-build/test_custom_ops 2024-08-06T20:39:04.4754981Z creating: build/custom_test_artifacts/jit-hook-build/ 2024-08-06T20:39:04.4755601Z creating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/ 2024-08-06T20:39:04.4756303Z creating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/pkgRedirects/ 2024-08-06T20:39:04.4764258Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/CMakeConfigureLog.yaml 2024-08-06T20:39:04.4765163Z creating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/ 2024-08-06T20:39:04.4766046Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CMakeSystem.cmake 2024-08-06T20:39:04.4767195Z creating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdC/ 2024-08-06T20:39:04.4768114Z creating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdC/tmp/ 2024-08-06T20:39:04.4769341Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdC/CMakeCCompilerId.c 2024-08-06T20:39:04.4770996Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdC/a.out 2024-08-06T20:39:04.4771889Z creating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCXX/ 2024-08-06T20:39:04.4772731Z creating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCXX/tmp/ 2024-08-06T20:39:04.4775184Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCXX/CMakeCXXCompilerId.cpp 2024-08-06T20:39:04.4776759Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCXX/a.out 2024-08-06T20:39:04.4779133Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CMakeDetermineCompilerABI_C.bin 2024-08-06T20:39:04.4780156Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CMakeCCompiler.cmake 2024-08-06T20:39:04.4781827Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CMakeDetermineCompilerABI_CXX.bin 2024-08-06T20:39:04.4783317Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CMakeCXXCompiler.cmake 2024-08-06T20:39:04.4784222Z creating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/ 2024-08-06T20:39:04.4785096Z creating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/ 2024-08-06T20:39:04.4832189Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cpp4.ii 2024-08-06T20:39:04.4878879Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cudafe1.cpp 2024-08-06T20:39:04.4880149Z extracting: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.module_id 2024-08-06T20:39:04.4933884Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cpp1.ii 2024-08-06T20:39:04.4935266Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cudafe1.c 2024-08-06T20:39:04.4936851Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cudafe1.gpu 2024-08-06T20:39:04.4938320Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cudafe1.stub.c 2024-08-06T20:39:04.4939703Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.ptx 2024-08-06T20:39:04.4941084Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.sm_52.cubin 2024-08-06T20:39:04.4942472Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.fatbin 2024-08-06T20:39:04.4943822Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.fatbin.c 2024-08-06T20:39:04.4945168Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.o 2024-08-06T20:39:04.4946442Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.sm_52.cubin 2024-08-06T20:39:04.4947650Z extracting: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.reg.c 2024-08-06T20:39:04.4948826Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.fatbin 2024-08-06T20:39:04.4950044Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.fatbin.c 2024-08-06T20:39:04.4951358Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.o 2024-08-06T20:39:04.4952537Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/CMakeCUDACompilerId.cu 2024-08-06T20:39:04.5021867Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CompilerIdCUDA/a.out 2024-08-06T20:39:04.5094690Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CMakeDetermineCompilerABI_CUDA.bin 2024-08-06T20:39:04.5095745Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/3.26.4/CMakeCUDACompiler.cmake 2024-08-06T20:39:04.5096702Z creating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/CMakeScratch/ 2024-08-06T20:39:04.5097531Z creating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/CMakeTmp/ 2024-08-06T20:39:04.5098398Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/cmake.check_cache 2024-08-06T20:39:04.5099319Z creating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/test_jit_hooks.dir/ 2024-08-06T20:39:04.5100366Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/test_jit_hooks.dir/compiler_depend.ts 2024-08-06T20:39:04.5101548Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/test_jit_hooks.dir/compiler_depend.make 2024-08-06T20:39:04.5102685Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/test_jit_hooks.dir/depend.make 2024-08-06T20:39:04.5103731Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/test_jit_hooks.dir/link.txt 2024-08-06T20:39:04.5104986Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/test_jit_hooks.dir/cmake_clean.cmake 2024-08-06T20:39:04.5105983Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/test_jit_hooks.dir/build.make 2024-08-06T20:39:04.5106972Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/test_jit_hooks.dir/DependInfo.cmake 2024-08-06T20:39:04.5107971Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/test_jit_hooks.dir/flags.make 2024-08-06T20:39:04.5108949Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/test_jit_hooks.dir/progress.make 2024-08-06T20:39:04.5130393Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/test_jit_hooks.dir/test_jit_hooks.cpp.o.d 2024-08-06T20:39:04.5207774Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/test_jit_hooks.dir/test_jit_hooks.cpp.o 2024-08-06T20:39:04.5208868Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/CMakeDirectoryInformation.cmake 2024-08-06T20:39:04.5209928Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/TargetDirectories.txt 2024-08-06T20:39:04.5210880Z extracting: build/custom_test_artifacts/jit-hook-build/CMakeFiles/progress.marks 2024-08-06T20:39:04.5211752Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/Makefile2 2024-08-06T20:39:04.5212637Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeFiles/Makefile.cmake 2024-08-06T20:39:04.5213515Z inflating: build/custom_test_artifacts/jit-hook-build/detect_cuda_version.cc 2024-08-06T20:39:04.5216610Z inflating: build/custom_test_artifacts/jit-hook-build/CMakeCache.txt 2024-08-06T20:39:04.5217319Z inflating: build/custom_test_artifacts/jit-hook-build/Makefile 2024-08-06T20:39:04.5218270Z inflating: build/custom_test_artifacts/jit-hook-build/cmake_install.cmake 2024-08-06T20:39:04.5278441Z inflating: build/custom_test_artifacts/jit-hook-build/test_jit_hooks 2024-08-06T20:39:04.5279097Z creating: build/custom_test_artifacts/custom-backend-build/ 2024-08-06T20:39:04.5279764Z creating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/ 2024-08-06T20:39:04.5280542Z creating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/pkgRedirects/ 2024-08-06T20:39:04.5288212Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/CMakeConfigureLog.yaml 2024-08-06T20:39:04.5289271Z creating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/ 2024-08-06T20:39:04.5290148Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CMakeSystem.cmake 2024-08-06T20:39:04.5291092Z creating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdC/ 2024-08-06T20:39:04.5291995Z creating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdC/tmp/ 2024-08-06T20:39:04.5293342Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdC/CMakeCCompilerId.c 2024-08-06T20:39:04.5295627Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdC/a.out 2024-08-06T20:39:04.5296579Z creating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCXX/ 2024-08-06T20:39:04.5297493Z creating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCXX/tmp/ 2024-08-06T20:39:04.5299125Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCXX/CMakeCXXCompilerId.cpp 2024-08-06T20:39:04.5301372Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCXX/a.out 2024-08-06T20:39:04.5303125Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CMakeDetermineCompilerABI_C.bin 2024-08-06T20:39:04.5304424Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CMakeCCompiler.cmake 2024-08-06T20:39:04.5306670Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CMakeDetermineCompilerABI_CXX.bin 2024-08-06T20:39:04.5307964Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CMakeCXXCompiler.cmake 2024-08-06T20:39:04.5308955Z creating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/ 2024-08-06T20:39:04.5309889Z creating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/ 2024-08-06T20:39:04.5356612Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cpp4.ii 2024-08-06T20:39:04.5403332Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cudafe1.cpp 2024-08-06T20:39:04.5405385Z extracting: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.module_id 2024-08-06T20:39:04.5458253Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cpp1.ii 2024-08-06T20:39:04.5459731Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cudafe1.c 2024-08-06T20:39:04.5461215Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cudafe1.gpu 2024-08-06T20:39:04.5462741Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.cudafe1.stub.c 2024-08-06T20:39:04.5464177Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.ptx 2024-08-06T20:39:04.5465629Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.sm_52.cubin 2024-08-06T20:39:04.5467097Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.fatbin 2024-08-06T20:39:04.5468526Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.fatbin.c 2024-08-06T20:39:04.5469943Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/CMakeCUDACompilerId.o 2024-08-06T20:39:04.5471445Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.sm_52.cubin 2024-08-06T20:39:04.5472742Z extracting: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.reg.c 2024-08-06T20:39:04.5474002Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.fatbin 2024-08-06T20:39:04.5475309Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.fatbin.c 2024-08-06T20:39:04.5476565Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/tmp/a_dlink.o 2024-08-06T20:39:04.5477827Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/CMakeCUDACompilerId.cu 2024-08-06T20:39:04.5546486Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CompilerIdCUDA/a.out 2024-08-06T20:39:04.5619560Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CMakeDetermineCompilerABI_CUDA.bin 2024-08-06T20:39:04.5620691Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/3.26.4/CMakeCUDACompiler.cmake 2024-08-06T20:39:04.5621637Z creating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/CMakeScratch/ 2024-08-06T20:39:04.5622441Z creating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/CMakeTmp/ 2024-08-06T20:39:04.5623307Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/cmake.check_cache 2024-08-06T20:39:04.5624204Z creating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/custom_backend.dir/ 2024-08-06T20:39:04.5625214Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/custom_backend.dir/compiler_depend.ts 2024-08-06T20:39:04.5626357Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/custom_backend.dir/compiler_depend.make 2024-08-06T20:39:04.5627472Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/custom_backend.dir/depend.make 2024-08-06T20:39:04.5628500Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/custom_backend.dir/link.txt 2024-08-06T20:39:04.5629542Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/custom_backend.dir/cmake_clean.cmake 2024-08-06T20:39:04.5630776Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/custom_backend.dir/build.make 2024-08-06T20:39:04.5631871Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/custom_backend.dir/DependInfo.cmake 2024-08-06T20:39:04.5632981Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/custom_backend.dir/flags.make 2024-08-06T20:39:04.5634149Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/custom_backend.dir/progress.make 2024-08-06T20:39:04.5635574Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/custom_backend.dir/custom_backend.cpp.o.d 2024-08-06T20:39:04.5786366Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/custom_backend.dir/custom_backend.cpp.o 2024-08-06T20:39:04.5787417Z creating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/test_custom_backend.dir/ 2024-08-06T20:39:04.5788489Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/test_custom_backend.dir/compiler_depend.ts 2024-08-06T20:39:04.5789682Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/test_custom_backend.dir/compiler_depend.make 2024-08-06T20:39:04.5790836Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/test_custom_backend.dir/depend.make 2024-08-06T20:39:04.5791903Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/test_custom_backend.dir/link.txt 2024-08-06T20:39:04.5793022Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/test_custom_backend.dir/cmake_clean.cmake 2024-08-06T20:39:04.5794319Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/test_custom_backend.dir/build.make 2024-08-06T20:39:04.5795430Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/test_custom_backend.dir/DependInfo.cmake 2024-08-06T20:39:04.5796563Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/test_custom_backend.dir/flags.make 2024-08-06T20:39:04.5797684Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/test_custom_backend.dir/progress.make 2024-08-06T20:39:04.5819614Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/test_custom_backend.dir/test_custom_backend.cpp.o.d 2024-08-06T20:39:04.5885923Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/test_custom_backend.dir/test_custom_backend.cpp.o 2024-08-06T20:39:04.5887110Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/CMakeDirectoryInformation.cmake 2024-08-06T20:39:04.5888164Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/TargetDirectories.txt 2024-08-06T20:39:04.5889094Z extracting: build/custom_test_artifacts/custom-backend-build/CMakeFiles/progress.marks 2024-08-06T20:39:04.5889943Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/Makefile2 2024-08-06T20:39:04.5891105Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeFiles/Makefile.cmake 2024-08-06T20:39:04.5892066Z inflating: build/custom_test_artifacts/custom-backend-build/detect_cuda_version.cc 2024-08-06T20:39:04.5895338Z inflating: build/custom_test_artifacts/custom-backend-build/CMakeCache.txt 2024-08-06T20:39:04.5896097Z inflating: build/custom_test_artifacts/custom-backend-build/Makefile 2024-08-06T20:39:04.5897100Z inflating: build/custom_test_artifacts/custom-backend-build/cmake_install.cmake 2024-08-06T20:39:04.6022079Z inflating: build/custom_test_artifacts/custom-backend-build/libcustom_backend.so 2024-08-06T20:39:04.6073631Z inflating: build/custom_test_artifacts/custom-backend-build/test_custom_backend 2024-08-06T20:39:04.6074249Z creating: build/lib/ 2024-08-06T20:39:04.6085876Z inflating: build/lib/libpthreadpool.a 2024-08-06T20:39:04.6095874Z inflating: build/lib/libcpuinfo.a 2024-08-06T20:39:04.6105854Z inflating: build/lib/libcpuinfo_internals.a 2024-08-06T20:39:04.6106770Z inflating: build/lib/libclog.a 2024-08-06T20:39:04.6214105Z inflating: build/lib/libprotobuf-lite.a 2024-08-06T20:39:04.6216714Z inflating: build/lib/libnnpack_reference_layers.a 2024-08-06T20:39:04.6763698Z inflating: build/lib/libprotobuf.a 2024-08-06T20:39:04.6842208Z inflating: build/lib/libgtest.a 2024-08-06T20:39:04.6851535Z inflating: build/lib/libittnotify.a 2024-08-06T20:39:04.6941966Z inflating: build/lib/libbenchmark.a 2024-08-06T20:39:04.6976284Z inflating: build/lib/libtensorpipe_uv.a 2024-08-06T20:39:04.7053328Z inflating: build/lib/libasmjit.a 2024-08-06T20:39:04.7074970Z inflating: build/lib/libnnpack.a 2024-08-06T20:39:04.7228379Z inflating: build/lib/libgloo.a 2024-08-06T20:39:04.7253508Z inflating: build/lib/libfmt.a 2024-08-06T20:39:04.7255682Z inflating: build/lib/libcaffe2_nvrtc.so 2024-08-06T20:39:04.7370057Z inflating: build/lib/libc10.so 2024-08-06T20:39:04.7370792Z inflating: build/lib/libfoxi_loader.a 2024-08-06T20:39:04.7373017Z inflating: build/lib/libtorch_global_deps.so 2024-08-06T20:39:04.7396603Z inflating: build/lib/libpytorch_qnnpack.a 2024-08-06T20:39:04.8002805Z inflating: build/lib/libprotoc.a 2024-08-06T20:39:04.8025037Z inflating: build/lib/libgmock.a 2024-08-06T20:39:04.8025802Z inflating: build/lib/libgtest_main.a 2024-08-06T20:39:04.8026850Z inflating: build/lib/libbenchmark_main.a 2024-08-06T20:39:04.8713859Z inflating: build/lib/libtensorpipe.a 2024-08-06T20:39:06.0835654Z inflating: build/lib/libdnnl.a 2024-08-06T20:39:06.0905477Z inflating: build/lib/libc10_cuda.so 2024-08-06T20:39:06.2404624Z inflating: build/lib/libfbgemm.a 2024-08-06T20:39:06.2405827Z inflating: build/lib/libgmock_main.a 2024-08-06T20:39:06.3022891Z inflating: build/lib/libkineto.a 2024-08-06T20:39:06.3329899Z inflating: build/lib/libtensorpipe_cuda.a 2024-08-06T20:39:06.3553153Z inflating: build/lib/libXNNPACK.a 2024-08-06T20:39:06.3603901Z inflating: build/lib/libonnx_proto.a 2024-08-06T20:39:06.4436186Z inflating: build/lib/libonnx.a 2024-08-06T20:39:06.4896368Z inflating: build/lib/libgloo_cuda.a 2024-08-06T20:39:09.4080207Z inflating: build/lib/libtorch_cpu.so 2024-08-06T20:39:09.4085105Z inflating: build/lib/libshm.so 2024-08-06T20:39:09.4091192Z inflating: build/lib/libunbox_lib.a 2024-08-06T20:39:11.8642052Z inflating: build/lib/libtorch_cuda.so 2024-08-06T20:39:11.8643724Z inflating: build/lib/libtorch.so 2024-08-06T20:39:11.8646695Z inflating: build/lib/libc10d_cuda_test.so 2024-08-06T20:39:12.8621336Z inflating: build/lib/libtorch_cuda_linalg.so 2024-08-06T20:39:12.8646225Z inflating: build/lib/libjitbackend_test.so 2024-08-06T20:39:12.8733087Z inflating: build/lib/libtorchbind_test.so 2024-08-06T20:39:12.8763967Z inflating: build/lib/libbackend_with_compiler.so 2024-08-06T20:39:12.8795505Z inflating: build/lib/libaoti_custom_ops.so 2024-08-06T20:39:13.1197394Z inflating: build/lib/libtorch_python.so 2024-08-06T20:39:13.1240175Z inflating: build/lib/libnnapi_backend.so 2024-08-06T20:39:13.1240616Z creating: build/bin/ 2024-08-06T20:39:13.1301639Z inflating: build/bin/c10_CompileTimeFunctionPointer_test 2024-08-06T20:39:13.1364356Z inflating: build/bin/c10_DeviceGuard_test 2024-08-06T20:39:13.1426083Z inflating: build/bin/c10_Device_test 2024-08-06T20:39:13.1497583Z inflating: build/bin/c10_DispatchKeySet_test 2024-08-06T20:39:13.1562882Z inflating: build/bin/c10_Scalar_test 2024-08-06T20:39:13.1622812Z inflating: build/bin/c10_StreamGuard_test 2024-08-06T20:39:13.1685027Z inflating: build/bin/c10_SymInt_test 2024-08-06T20:39:13.1751862Z inflating: build/bin/c10_InlineDeviceGuard_test 2024-08-06T20:39:13.1820389Z inflating: build/bin/c10_InlineStreamGuard_test 2024-08-06T20:39:13.1889059Z inflating: build/bin/c10_SizesAndStrides_test 2024-08-06T20:39:13.1975829Z inflating: build/bin/c10_cow_test 2024-08-06T20:39:13.2040476Z inflating: build/bin/c10_Bitset_test 2024-08-06T20:39:13.2100429Z inflating: build/bin/c10_ConstexprCrc_test 2024-08-06T20:39:13.2161357Z inflating: build/bin/c10_DeadlockDetection_test 2024-08-06T20:39:13.2223697Z inflating: build/bin/c10_Half_test 2024-08-06T20:39:13.2292022Z inflating: build/bin/c10_LeftRight_test 2024-08-06T20:39:13.2359350Z inflating: build/bin/c10_Metaprogramming_test 2024-08-06T20:39:13.2420303Z inflating: build/bin/c10_Synchronized_test 2024-08-06T20:39:13.2487987Z inflating: build/bin/c10_ThreadLocal_test 2024-08-06T20:39:13.2550908Z inflating: build/bin/c10_TypeIndex_test 2024-08-06T20:39:13.2613376Z inflating: build/bin/c10_TypeList_test 2024-08-06T20:39:13.2672695Z inflating: build/bin/c10_TypeTraits_test 2024-08-06T20:39:13.2735949Z inflating: build/bin/c10_accumulate_test 2024-08-06T20:39:13.2803477Z inflating: build/bin/c10_bfloat16_test 2024-08-06T20:39:13.2865237Z inflating: build/bin/c10_bit_cast_test 2024-08-06T20:39:13.2934032Z inflating: build/bin/c10_complex_math_test 2024-08-06T20:39:13.3001580Z inflating: build/bin/c10_complex_test 2024-08-06T20:39:13.3065998Z inflating: build/bin/c10_exception_test 2024-08-06T20:39:13.3127633Z inflating: build/bin/c10_flags_test 2024-08-06T20:39:13.3188351Z inflating: build/bin/c10_generic_math_test 2024-08-06T20:39:13.3386702Z inflating: build/bin/c10_intrusive_ptr_test 2024-08-06T20:39:13.3448848Z inflating: build/bin/c10_irange_test 2024-08-06T20:39:13.3513932Z inflating: build/bin/c10_lazy_test 2024-08-06T20:39:13.3583374Z inflating: build/bin/c10_logging_test 2024-08-06T20:39:13.3674266Z inflating: build/bin/c10_optional_test 2024-08-06T20:39:13.3750535Z inflating: build/bin/c10_ordered_preserving_dict_test 2024-08-06T20:39:13.3816246Z inflating: build/bin/c10_registry_test 2024-08-06T20:39:13.3879104Z inflating: build/bin/c10_ssize_test 2024-08-06T20:39:13.4061483Z inflating: build/bin/c10_small_vector_test 2024-08-06T20:39:13.4124718Z inflating: build/bin/c10_string_util_test 2024-08-06T20:39:13.4196310Z inflating: build/bin/c10_string_view_test 2024-08-06T20:39:13.4257802Z inflating: build/bin/c10_tempfile_test 2024-08-06T20:39:13.4326248Z inflating: build/bin/c10_typeid_test 2024-08-06T20:39:13.4385410Z inflating: build/bin/c10_intrusive_ptr_benchmark 2024-08-06T20:39:13.4927685Z inflating: build/bin/protoc-3.13.0.0 2024-08-06T20:39:13.5469578Z inflating: build/bin/protoc 2024-08-06T20:39:13.5534278Z inflating: build/bin/c10_cuda_CUDAAssertionsTest_1_var_test 2024-08-06T20:39:13.5598417Z inflating: build/bin/c10_cuda_CUDAAssertionsTest_catches_stream 2024-08-06T20:39:13.5662839Z inflating: build/bin/c10_cuda_CUDAAssertionsTest_catches_thread_and_block_and_device 2024-08-06T20:39:13.5726243Z inflating: build/bin/c10_cuda_CUDAAssertionsTest_from_2_processes 2024-08-06T20:39:13.5790432Z inflating: build/bin/c10_cuda_CUDAAssertionsTest_multiple_writes_from_blocks_and_threads 2024-08-06T20:39:13.5854876Z inflating: build/bin/c10_cuda_CUDAAssertionsTest_multiple_writes_from_multiple_blocks 2024-08-06T20:39:13.5919169Z inflating: build/bin/c10_cuda_CUDAAssertionsTest_multiple_writes_from_same_block 2024-08-06T20:39:13.5978837Z inflating: build/bin/c10_cuda_CUDATest 2024-08-06T20:39:13.6376007Z inflating: build/bin/vec_test_all_types_DEFAULT 2024-08-06T20:39:13.6809384Z inflating: build/bin/vec_test_all_types_AVX2 2024-08-06T20:39:13.7223196Z inflating: build/bin/vec_test_all_types_AVX512 2024-08-06T20:39:13.7288373Z inflating: build/bin/HashStoreTest 2024-08-06T20:39:13.7353071Z inflating: build/bin/FileStoreTest 2024-08-06T20:39:13.7421013Z inflating: build/bin/TCPStoreTest 2024-08-06T20:39:13.7484938Z inflating: build/bin/BackoffTest 2024-08-06T20:39:13.7502323Z inflating: build/bin/ProcessGroupMPITest 2024-08-06T20:39:13.7507942Z inflating: build/bin/torch_shm_manager 2024-08-06T20:39:13.7573428Z inflating: build/bin/test_edge_op_registration 2024-08-06T20:39:13.7577322Z inflating: build/bin/example_allreduce 2024-08-06T20:39:13.7640678Z inflating: build/bin/Dimname_test 2024-08-06T20:39:13.7730300Z inflating: build/bin/Dict_test 2024-08-06T20:39:13.7808781Z inflating: build/bin/MaybeOwned_test 2024-08-06T20:39:13.7880457Z inflating: build/bin/apply_utils_test 2024-08-06T20:39:13.7950774Z inflating: build/bin/NamedTensor_test 2024-08-06T20:39:13.8022767Z inflating: build/bin/atest 2024-08-06T20:39:13.8099349Z inflating: build/bin/basic 2024-08-06T20:39:13.8166822Z inflating: build/bin/broadcast_test 2024-08-06T20:39:13.8229236Z inflating: build/bin/cpu_allocator_test 2024-08-06T20:39:13.8299859Z inflating: build/bin/cpu_generator_test 2024-08-06T20:39:13.8364934Z inflating: build/bin/cpu_profiling_allocator_test 2024-08-06T20:39:13.8426311Z inflating: build/bin/dispatch_key_set_test 2024-08-06T20:39:13.8539456Z inflating: build/bin/cpu_rng_test 2024-08-06T20:39:13.8600508Z inflating: build/bin/dlconvertor_test 2024-08-06T20:39:13.8671852Z inflating: build/bin/extension_backend_test 2024-08-06T20:39:13.8739292Z inflating: build/bin/half_test 2024-08-06T20:39:13.8855971Z inflating: build/bin/ivalue_test 2024-08-06T20:39:13.8917015Z inflating: build/bin/lazy_tensor_test 2024-08-06T20:39:13.8983264Z inflating: build/bin/math_kernel_test 2024-08-06T20:39:13.9049325Z inflating: build/bin/memory_format_test 2024-08-06T20:39:13.9114281Z inflating: build/bin/memory_overlapping_test 2024-08-06T20:39:13.9178611Z inflating: build/bin/mobile_memory_cleanup 2024-08-06T20:39:13.9247270Z inflating: build/bin/native_test 2024-08-06T20:39:13.9309305Z inflating: build/bin/operator_name_test 2024-08-06T20:39:13.9372158Z inflating: build/bin/packedtensoraccessor_test 2024-08-06T20:39:13.9435452Z inflating: build/bin/operators_test 2024-08-06T20:39:13.9517365Z inflating: build/bin/pow_test 2024-08-06T20:39:13.9587067Z inflating: build/bin/quantized_test 2024-08-06T20:39:13.9648421Z inflating: build/bin/reduce_ops_test 2024-08-06T20:39:13.9711018Z inflating: build/bin/reportMemoryUsage_test 2024-08-06T20:39:13.9779610Z inflating: build/bin/scalar_tensor_test 2024-08-06T20:39:13.9843976Z inflating: build/bin/StorageUtils_test 2024-08-06T20:39:13.9915643Z inflating: build/bin/scalar_test 2024-08-06T20:39:13.9979537Z inflating: build/bin/stride_properties_test 2024-08-06T20:39:14.0045973Z inflating: build/bin/test_parallel 2024-08-06T20:39:14.0141715Z inflating: build/bin/tensor_iterator_test 2024-08-06T20:39:14.0209883Z inflating: build/bin/type_ptr_test 2024-08-06T20:39:14.0213596Z inflating: build/bin/thread_init_test 2024-08-06T20:39:14.0286296Z inflating: build/bin/type_test 2024-08-06T20:39:14.0288709Z inflating: build/bin/verify_api_visibility 2024-08-06T20:39:14.0352869Z inflating: build/bin/undefined_tensor_test 2024-08-06T20:39:14.0437548Z inflating: build/bin/legacy_vmap_test 2024-08-06T20:39:14.0500166Z inflating: build/bin/weakref_test 2024-08-06T20:39:14.0563385Z inflating: build/bin/wrapdim_test 2024-08-06T20:39:14.0626869Z inflating: build/bin/xla_tensor_test 2024-08-06T20:39:14.0755508Z inflating: build/bin/List_test 2024-08-06T20:39:14.0829016Z inflating: build/bin/IListRef_test 2024-08-06T20:39:14.0976869Z inflating: build/bin/kernel_function_legacy_test 2024-08-06T20:39:14.1058074Z inflating: build/bin/KernelFunction_test 2024-08-06T20:39:14.1175410Z inflating: build/bin/kernel_function_test 2024-08-06T20:39:14.1331724Z inflating: build/bin/kernel_lambda_legacy_test 2024-08-06T20:39:14.1457855Z inflating: build/bin/kernel_lambda_test 2024-08-06T20:39:14.1575300Z inflating: build/bin/make_boxed_from_unboxed_functor_test 2024-08-06T20:39:14.1650161Z inflating: build/bin/kernel_stackbased_test 2024-08-06T20:39:14.1712499Z inflating: build/bin/CppSignature_test 2024-08-06T20:39:14.1771931Z inflating: build/bin/op_allowlist_test 2024-08-06T20:39:14.1840689Z inflating: build/bin/backend_fallback_test 2024-08-06T20:39:14.1918371Z inflating: build/bin/inline_container_test 2024-08-06T20:39:14.2282768Z inflating: build/bin/op_registration_test 2024-08-06T20:39:14.2347599Z inflating: build/bin/cuda_apply_test 2024-08-06T20:39:14.2411960Z inflating: build/bin/cuda_allocator_test 2024-08-06T20:39:14.2483705Z inflating: build/bin/cuda_atomic_ops_test 2024-08-06T20:39:14.2550822Z inflating: build/bin/cuda_caching_host_allocator_test 2024-08-06T20:39:14.2636638Z inflating: build/bin/cuda_complex_math_test 2024-08-06T20:39:14.2696906Z inflating: build/bin/cuda_device_test 2024-08-06T20:39:14.2769407Z inflating: build/bin/cuda_complex_test 2024-08-06T20:39:14.2839757Z inflating: build/bin/cuda_cub_test 2024-08-06T20:39:14.2901580Z inflating: build/bin/cuda_dlconvertor_test 2024-08-06T20:39:14.2964253Z inflating: build/bin/cuda_integer_divider_test 2024-08-06T20:39:14.3044759Z inflating: build/bin/cuda_distributions_test 2024-08-06T20:39:14.3105623Z inflating: build/bin/cuda_half_test 2024-08-06T20:39:14.3174605Z inflating: build/bin/cuda_generator_test 2024-08-06T20:39:14.3235177Z inflating: build/bin/cuda_optional_test 2024-08-06T20:39:14.3299203Z inflating: build/bin/cuda_reportMemoryUsage_test 2024-08-06T20:39:14.3360451Z inflating: build/bin/cuda_allocatorTraceTracker_test 2024-08-06T20:39:14.3434672Z inflating: build/bin/cuda_stream_test 2024-08-06T20:39:14.3497552Z inflating: build/bin/cuda_packedtensoraccessor_test 2024-08-06T20:39:14.3558052Z inflating: build/bin/cuda_cudnn_test 2024-08-06T20:39:14.3622069Z inflating: build/bin/cuda_vectorized_test 2024-08-06T20:39:14.3639616Z inflating: build/bin/tutorial_tensorexpr 2024-08-06T20:39:14.3720054Z inflating: build/bin/ProcessGroupGlooTest 2024-08-06T20:39:14.3790164Z inflating: build/bin/ProcessGroupGlooAsyncTest 2024-08-06T20:39:14.3867980Z inflating: build/bin/ProcessGroupNCCLTest 2024-08-06T20:39:14.3943973Z inflating: build/bin/ProcessGroupNCCLErrorsTest 2024-08-06T20:39:14.4011339Z inflating: build/bin/test_dist_autograd 2024-08-06T20:39:14.4014572Z inflating: build/bin/parallel_benchmark 2024-08-06T20:39:14.4025533Z inflating: build/bin/aot_model_compiler_test 2024-08-06T20:39:14.4109978Z inflating: build/bin/test_cpp_rpc 2024-08-06T20:39:14.4191534Z inflating: build/bin/test_mobile_nnc 2024-08-06T20:39:14.4610392Z inflating: build/bin/test_lazy 2024-08-06T20:39:14.5612712Z inflating: build/bin/test_tensorexpr 2024-08-06T20:39:14.7041705Z inflating: build/bin/test_api 2024-08-06T20:39:14.7744483Z inflating: build/bin/test_jit 2024-08-06T20:39:14.7744904Z creating: .additional_ci_files/ 2024-08-06T20:39:14.7814623Z inflating: .additional_ci_files/test-times.json 2024-08-06T20:39:14.8092522Z inflating: .additional_ci_files/test-class-times.json 2024-08-06T20:39:14.8143414Z ##[group]Run rm artifacts.zip 2024-08-06T20:39:14.8143828Z rm artifacts.zip 2024-08-06T20:39:14.8151113Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:39:14.8151600Z env: 2024-08-06T20:39:14.8151891Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:14.8152303Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:14.8152764Z ##[endgroup] 2024-08-06T20:39:14.8907432Z ##[group]Run df -H 2024-08-06T20:39:14.8907766Z df -H 2024-08-06T20:39:14.8914160Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:39:14.8914637Z env: 2024-08-06T20:39:14.8914926Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:14.8915355Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:14.8915815Z ##[endgroup] 2024-08-06T20:39:14.8959650Z Filesystem Size Used Avail Use% Mounted on 2024-08-06T20:39:14.8960121Z devtmpfs 4.2M 0 4.2M 0% /dev 2024-08-06T20:39:14.8960556Z tmpfs 65G 0 65G 0% /dev/shm 2024-08-06T20:39:14.8960987Z tmpfs 26G 562k 26G 1% /run 2024-08-06T20:39:14.8961592Z /dev/xvda1 161G 38G 124G 24% / 2024-08-06T20:39:14.8962276Z tmpfs 65G 21k 65G 1% /tmp 2024-08-06T20:39:14.8962709Z /dev/xvda128 11M 1.4M 9.2M 13% /boot/efi 2024-08-06T20:39:14.8963145Z tmpfs 13G 0 13G 0% /run/user/0 2024-08-06T20:39:14.9002558Z Prepare all required actions 2024-08-06T20:39:14.9003074Z Getting action download info 2024-08-06T20:39:15.0040768Z ##[group]Run ./.github/actions/download-td-artifacts 2024-08-06T20:39:15.0041229Z with: 2024-08-06T20:39:15.0041504Z env: 2024-08-06T20:39:15.0041796Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:15.0042235Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:15.0042679Z ##[endgroup] 2024-08-06T20:39:15.0108009Z ##[group]Run seemethere/download-artifact-s3@v4 2024-08-06T20:39:15.0108449Z with: 2024-08-06T20:39:15.0108741Z name: td_results 2024-08-06T20:39:15.0109045Z s3-bucket: gha-artifacts 2024-08-06T20:39:15.0109397Z region: us-east-1 2024-08-06T20:39:15.0109701Z env: 2024-08-06T20:39:15.0109968Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:15.0110396Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:15.0110858Z ##[endgroup] 2024-08-06T20:39:15.6169679Z (node:205364) NOTE: We are formalizing our plans to enter AWS SDK for JavaScript (v2) into maintenance mode in 2023. 2024-08-06T20:39:15.6170302Z 2024-08-06T20:39:15.6170568Z Please migrate your code to use AWS SDK for JavaScript (v3). 2024-08-06T20:39:15.6171218Z For more information, check the migration guide at https://a.co/7PzMCcy 2024-08-06T20:39:15.6172183Z (Use `node --trace-warnings ...` to show where the warning was created) 2024-08-06T20:39:15.6987233Z Found 1 objects with prefix pytorch/pytorch/10273124344/td_results/ 2024-08-06T20:39:15.6988028Z Starting download (1/1): /home/ec2-user/actions-runner/_work/pytorch/pytorch/td_results.json 2024-08-06T20:39:15.7569664Z Finished download (1/1): /home/ec2-user/actions-runner/_work/pytorch/pytorch/td_results.json 2024-08-06T20:39:15.7580256Z Artifact download has finished successfully 2024-08-06T20:39:15.7801706Z ##[group]Run mkdir -p .additional_ci_files 2024-08-06T20:39:15.7802166Z mkdir -p .additional_ci_files 2024-08-06T20:39:15.7802700Z mv td_results.json .additional_ci_files/td_results.json 2024-08-06T20:39:15.7811316Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:39:15.7811808Z env: 2024-08-06T20:39:15.7812099Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:15.7812513Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:15.7812998Z ##[endgroup] 2024-08-06T20:39:15.7916270Z ##[group]Run .github/scripts/parse_ref.py 2024-08-06T20:39:15.7916780Z .github/scripts/parse_ref.py 2024-08-06T20:39:15.7923248Z shell: /usr/bin/bash -e {0} 2024-08-06T20:39:15.7923589Z env: 2024-08-06T20:39:15.7923874Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:15.7924306Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:15.7924753Z ##[endgroup] 2024-08-06T20:39:15.8267537Z Prepare all required actions 2024-08-06T20:39:15.8316795Z ##[group]Run ./.github/actions/get-workflow-job-id 2024-08-06T20:39:15.8317228Z with: 2024-08-06T20:39:15.8317741Z github-token: *** 2024-08-06T20:39:15.8318057Z env: 2024-08-06T20:39:15.8318338Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:15.8318751Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:15.8319211Z ##[endgroup] 2024-08-06T20:39:15.8347068Z ##[group]Run set -eux 2024-08-06T20:39:15.8347423Z set -eux 2024-08-06T20:39:15.8347977Z python3 .github/scripts/get_workflow_job_id.py "${GITHUB_RUN_ID}" "${RUNNER_NAME}" 2024-08-06T20:39:15.8354396Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:39:15.8354882Z env: 2024-08-06T20:39:15.8355154Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:15.8355585Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:15.8356241Z GITHUB_TOKEN: *** 2024-08-06T20:39:15.8356545Z ##[endgroup] 2024-08-06T20:39:15.8382673Z + python3 .github/scripts/get_workflow_job_id.py 10273124344 i-0de4eefb8fab684da 2024-08-06T20:39:19.2667429Z setting job-id=28428065757 2024-08-06T20:39:19.2668162Z setting job-name=linux-focal-cuda12.1-py3.10-gcc9 / test (default, 1, 5, amz2023.linux.4xlarge.nvidia.gpu) 2024-08-06T20:39:19.2886458Z Prepare all required actions 2024-08-06T20:39:19.2886989Z Getting action download info 2024-08-06T20:39:19.4170683Z ##[group]Run ./.github/actions/filter-test-configs 2024-08-06T20:39:19.4171138Z with: 2024-08-06T20:39:19.4171656Z github-token: *** 2024-08-06T20:39:19.4173712Z test-matrix: {"include": [{"config": "default", "shard": 1, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 2, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 3, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 4, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 5, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}]} 2024-08-06T20:39:19.4176115Z job-name: linux-focal-cuda12.1-py3.10-gcc9 / test (default, 1, 5, amz2023.linux.4xlarge.nvidia.gpu) 2024-08-06T20:39:19.4176828Z env: 2024-08-06T20:39:19.4177180Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:19.4177619Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:19.4178078Z ##[endgroup] 2024-08-06T20:39:19.4232145Z ##[group]Run nick-fields/retry@3e91a01664abd3c5cd539100d10d33b9c5b68482 2024-08-06T20:39:19.4232961Z with: 2024-08-06T20:39:19.4233228Z shell: bash 2024-08-06T20:39:19.4233532Z timeout_minutes: 10 2024-08-06T20:39:19.4233845Z max_attempts: 5 2024-08-06T20:39:19.4234167Z retry_wait_seconds: 30 2024-08-06T20:39:19.4235130Z command: set -eux # PyYAML 6.0 doesn't work with MacOS x86 anymore # This must run on Python-3.7 (AmazonLinux2) so can't use request=3.32.2 python3 -m pip install requests==2.27.1 pyyaml==6.0.1 2024-08-06T20:39:19.4236147Z polling_interval_seconds: 1 2024-08-06T20:39:19.4236502Z warning_on_retry: true 2024-08-06T20:39:19.4236854Z continue_on_error: false 2024-08-06T20:39:19.4237197Z env: 2024-08-06T20:39:19.4237463Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:19.4237904Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:19.4238747Z GITHUB_TOKEN: *** 2024-08-06T20:39:19.4239058Z ##[endgroup] 2024-08-06T20:39:19.5156818Z + python3 -m pip install requests==2.27.1 pyyaml==6.0.1 2024-08-06T20:39:19.8425754Z Defaulting to user installation because normal site-packages is not writeable 2024-08-06T20:39:19.8629237Z Requirement already satisfied: requests==2.27.1 in /home/ec2-user/.local/lib/python3.9/site-packages (2.27.1) 2024-08-06T20:39:19.8635232Z Requirement already satisfied: pyyaml==6.0.1 in /home/ec2-user/.local/lib/python3.9/site-packages (6.0.1) 2024-08-06T20:39:19.8799264Z Requirement already satisfied: idna<4,>=2.5 in /usr/lib/python3.9/site-packages (from requests==2.27.1) (2.10) 2024-08-06T20:39:19.8805319Z Requirement already satisfied: certifi>=2017.4.17 in /home/ec2-user/.local/lib/python3.9/site-packages (from requests==2.27.1) (2024.7.4) 2024-08-06T20:39:19.8817819Z Requirement already satisfied: charset-normalizer~=2.0.0 in /home/ec2-user/.local/lib/python3.9/site-packages (from requests==2.27.1) (2.0.12) 2024-08-06T20:39:19.8825556Z Requirement already satisfied: urllib3<1.27,>=1.21.1 in /usr/lib/python3.9/site-packages (from requests==2.27.1) (1.25.10) 2024-08-06T20:39:20.4961863Z Command completed after 1 attempt(s). 2024-08-06T20:39:20.5022705Z ##[group]Run set -x 2024-08-06T20:39:20.5023053Z set -x 2024-08-06T20:39:20.5023332Z  2024-08-06T20:39:20.5023817Z # Use relative path here as this could be checked out anywhere, not necessarily 2024-08-06T20:39:20.5024415Z # in runner workspace 2024-08-06T20:39:20.5024890Z python3 "${GITHUB_ACTION_PATH}/../../scripts/parse_ref.py" 2024-08-06T20:39:20.5033841Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:39:20.5034322Z env: 2024-08-06T20:39:20.5034591Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:20.5035021Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:20.5035476Z ##[endgroup] 2024-08-06T20:39:20.5065681Z + python3 /home/ec2-user/actions-runner/_work/pytorch/pytorch/./.github/actions/filter-test-configs/../../scripts/parse_ref.py 2024-08-06T20:39:20.5381883Z ##[group]Run echo "Workflow: ${GITHUB_WORKFLOW}" 2024-08-06T20:39:20.5382410Z echo "Workflow: ${GITHUB_WORKFLOW}" 2024-08-06T20:39:20.5382848Z echo "Job name: ${JOB_NAME}" 2024-08-06T20:39:20.5383236Z  2024-08-06T20:39:20.5383720Z # Use relative path here as this could be checked out anywhere, not necessarily 2024-08-06T20:39:20.5384304Z # in runner workspace 2024-08-06T20:39:20.5384852Z python3 "${GITHUB_ACTION_PATH}/../../scripts/filter_test_configs.py" \ 2024-08-06T20:39:20.5385458Z  --workflow "${GITHUB_WORKFLOW}" \ 2024-08-06T20:39:20.5385879Z  --job-name "${JOB_NAME}" \ 2024-08-06T20:39:20.5388054Z  --test-matrix "{"include": [{"config": "default", "shard": 1, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 2, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 3, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 4, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 5, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}]}" \ 2024-08-06T20:39:20.5390459Z  --selected-test-configs "" \ 2024-08-06T20:39:20.5390903Z  --pr-number "${PR_NUMBER}" \ 2024-08-06T20:39:20.5391311Z  --tag "${TAG}" \ 2024-08-06T20:39:20.5391681Z  --event-name "${EVENT_NAME}" \ 2024-08-06T20:39:20.5392105Z  --schedule "${SCHEDULE}" \ 2024-08-06T20:39:20.5392510Z  --branch "${HEAD_BRANCH}" 2024-08-06T20:39:20.5398808Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:39:20.5399288Z env: 2024-08-06T20:39:20.5399564Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:20.5400012Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:20.5400702Z GITHUB_TOKEN: *** 2024-08-06T20:39:20.5401305Z JOB_NAME: linux-focal-cuda12.1-py3.10-gcc9 / test (default, 1, 5, amz2023.linux.4xlarge.nvidia.gpu) 2024-08-06T20:39:20.5401957Z PR_NUMBER: 132710 2024-08-06T20:39:20.5402281Z TAG: 2024-08-06T20:39:20.5402568Z EVENT_NAME: pull_request 2024-08-06T20:39:20.5402895Z SCHEDULE: 2024-08-06T20:39:20.5403185Z HEAD_BRANCH: 2024-08-06T20:39:20.5403485Z ##[endgroup] 2024-08-06T20:39:20.5430261Z Workflow: pull 2024-08-06T20:39:20.5431529Z Job name: linux-focal-cuda12.1-py3.10-gcc9 / test (default, 1, 5, amz2023.linux.4xlarge.nvidia.gpu) 2024-08-06T20:39:20.7988554Z INFO:root:Found no test-config label on the PR, so all test configs are included 2024-08-06T20:39:20.9910368Z ##[group]Run echo "Filtered matrix:" 2024-08-06T20:39:20.9910812Z echo "Filtered matrix:" 2024-08-06T20:39:20.9912966Z echo "{"include": [{"config": "default", "shard": 1, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 2, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 3, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 4, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}, {"config": "default", "shard": 5, "num_shards": 5, "runner": "amz2023.linux.4xlarge.nvidia.gpu"}]}" 2024-08-06T20:39:20.9915097Z  2024-08-06T20:39:20.9915372Z echo 2024-08-06T20:39:20.9915714Z echo "Is the current job unstable? False" 2024-08-06T20:39:20.9916236Z  2024-08-06T20:39:20.9916657Z echo 2024-08-06T20:39:20.9917252Z echo "Is keep-going label set? False" 2024-08-06T20:39:20.9917799Z  2024-08-06T20:39:20.9918183Z echo 2024-08-06T20:39:20.9918644Z echo "Renabled issues? " 2024-08-06T20:39:20.9925815Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:39:20.9926530Z env: 2024-08-06T20:39:20.9941775Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:20.9942338Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:20.9942819Z ##[endgroup] 2024-08-06T20:39:20.9972397Z Filtered matrix: 2024-08-06T20:39:20.9975086Z {include: [{config: default, shard: 1, num_shards: 5, runner: amz2023.linux.4xlarge.nvidia.gpu}, {config: default, shard: 2, num_shards: 5, runner: amz2023.linux.4xlarge.nvidia.gpu}, {config: default, shard: 3, num_shards: 5, runner: amz2023.linux.4xlarge.nvidia.gpu}, {config: default, shard: 4, num_shards: 5, runner: amz2023.linux.4xlarge.nvidia.gpu}, {config: default, shard: 5, num_shards: 5, runner: amz2023.linux.4xlarge.nvidia.gpu}]} 2024-08-06T20:39:20.9977582Z 2024-08-06T20:39:20.9977748Z Is the current job unstable? False 2024-08-06T20:39:20.9978244Z 2024-08-06T20:39:20.9978405Z Is keep-going label set? False 2024-08-06T20:39:20.9978653Z 2024-08-06T20:39:20.9978797Z Renabled issues? 2024-08-06T20:39:21.0035193Z ##[group]Run echo "timeout=$((JOB_TIMEOUT-30))" >> "${GITHUB_OUTPUT}" 2024-08-06T20:39:21.0035862Z echo "timeout=$((JOB_TIMEOUT-30))" >> "${GITHUB_OUTPUT}" 2024-08-06T20:39:21.0042071Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T20:39:21.0042553Z env: 2024-08-06T20:39:21.0043020Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:21.0043456Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:21.0043917Z JOB_TIMEOUT: 360 2024-08-06T20:39:21.0044209Z ##[endgroup] 2024-08-06T20:39:21.0130189Z ##[group]Run set -x 2024-08-06T20:39:21.0130634Z set -x 2024-08-06T20:39:21.0130918Z  2024-08-06T20:39:21.0131252Z if [[ $TEST_CONFIG == 'multigpu' ]]; then 2024-08-06T20:39:21.0131764Z  TEST_COMMAND=.ci/pytorch/multigpu-test.sh 2024-08-06T20:39:21.0132263Z elif [[ $BUILD_ENVIRONMENT == *onnx* ]]; then 2024-08-06T20:39:21.0132728Z  TEST_COMMAND=.ci/onnx/test.sh 2024-08-06T20:39:21.0133118Z else 2024-08-06T20:39:21.0133437Z  TEST_COMMAND=.ci/pytorch/test.sh 2024-08-06T20:39:21.0133838Z fi 2024-08-06T20:39:21.0134118Z  2024-08-06T20:39:21.0134541Z # detached container should get cleaned up by teardown_ec2_linux 2024-08-06T20:39:21.0135208Z # TODO: Stop building test binaries as part of the build phase 2024-08-06T20:39:21.0135813Z # Used for GPU_FLAG since that doesn't play nice 2024-08-06T20:39:21.0136345Z # shellcheck disable=SC2086,SC2090 2024-08-06T20:39:21.0136767Z container_name=$(docker run \ 2024-08-06T20:39:21.0137169Z  ${GPU_FLAG:-} \ 2024-08-06T20:39:21.0137539Z  -e BUILD_ENVIRONMENT \ 2024-08-06T20:39:21.0137906Z  -e PR_NUMBER \ 2024-08-06T20:39:21.0138258Z  -e GITHUB_ACTIONS \ 2024-08-06T20:39:21.0138635Z  -e GITHUB_REPOSITORY \ 2024-08-06T20:39:21.0139006Z  -e GITHUB_WORKFLOW \ 2024-08-06T20:39:21.0139377Z  -e GITHUB_JOB \ 2024-08-06T20:39:21.0139730Z  -e GITHUB_RUN_ID \ 2024-08-06T20:39:21.0140078Z  -e GITHUB_RUN_NUMBER \ 2024-08-06T20:39:21.0140461Z  -e GITHUB_RUN_ATTEMPT \ 2024-08-06T20:39:21.0140838Z  -e JOB_ID \ 2024-08-06T20:39:21.0141153Z  -e JOB_NAME \ 2024-08-06T20:39:21.0141504Z  -e BASE_SHA \ 2024-08-06T20:39:21.0141843Z  -e BRANCH \ 2024-08-06T20:39:21.0142154Z  -e SHA1 \ 2024-08-06T20:39:21.0142492Z  -e AWS_DEFAULT_REGION \ 2024-08-06T20:39:21.0142878Z  -e IN_WHEEL_TEST \ 2024-08-06T20:39:21.0143223Z  -e SHARD_NUMBER \ 2024-08-06T20:39:21.0143581Z  -e TEST_CONFIG \ 2024-08-06T20:39:21.0143943Z  -e NUM_TEST_SHARDS \ 2024-08-06T20:39:21.0144308Z  -e REENABLED_ISSUES \ 2024-08-06T20:39:21.0144702Z  -e CONTINUE_THROUGH_ERROR \ 2024-08-06T20:39:21.0145111Z  -e VERBOSE_TEST_LOGS \ 2024-08-06T20:39:21.0145480Z  -e TEST_SHOWLOCALS \ 2024-08-06T20:39:21.0145859Z  -e NO_TEST_TIMEOUT \ 2024-08-06T20:39:21.0146225Z  -e NO_TD \ 2024-08-06T20:39:21.0146541Z  -e TD_DISTRIBUTED \ 2024-08-06T20:39:21.0146908Z  -e PR_LABELS \ 2024-08-06T20:39:21.0147288Z  -e MAX_JOBS="$(nproc --ignore=2)" \ 2024-08-06T20:39:21.0147705Z  -e SCCACHE_BUCKET \ 2024-08-06T20:39:21.0148080Z  -e SCCACHE_S3_KEY_PREFIX \ 2024-08-06T20:39:21.0148464Z  -e XLA_CUDA \ 2024-08-06T20:39:21.0148826Z  -e XLA_CLANG_CACHE_S3_BUCKET_NAME \ 2024-08-06T20:39:21.0149285Z  -e PYTORCH_TEST_CUDA_MEM_LEAK_CHECK \ 2024-08-06T20:39:21.0149760Z  -e PYTORCH_TEST_RERUN_DISABLED_TESTS \ 2024-08-06T20:39:21.0150210Z  -e SKIP_SCCACHE_INITIALIZATION=1 \ 2024-08-06T20:39:21.0150648Z  -e HUGGING_FACE_HUB_TOKEN \ 2024-08-06T20:39:21.0151050Z  -e DASHBOARD_TAG \ 2024-08-06T20:39:21.0151468Z  --env-file="/tmp/github_env_${GITHUB_RUN_ID}" \ 2024-08-06T20:39:21.0151968Z  --security-opt seccomp=unconfined \ 2024-08-06T20:39:21.0152416Z  --cap-add=SYS_PTRACE \ 2024-08-06T20:39:21.0152780Z  --ipc=host \ 2024-08-06T20:39:21.0153127Z  --shm-size="${SHM_SIZE}" \ 2024-08-06T20:39:21.0153688Z  --tty \ 2024-08-06T20:39:21.0153985Z  --detach \ 2024-08-06T20:39:21.0154332Z  --name="${container_name}" \ 2024-08-06T20:39:21.0154849Z  --user jenkins \ 2024-08-06T20:39:21.0155298Z  -v "${GITHUB_WORKSPACE}:/var/lib/jenkins/workspace" \ 2024-08-06T20:39:21.0155802Z  -w /var/lib/jenkins/workspace \ 2024-08-06T20:39:21.0156216Z  "${DOCKER_IMAGE}" 2024-08-06T20:39:21.0156539Z ) 2024-08-06T20:39:21.0156912Z # Propagate download.pytorch.org IP to container 2024-08-06T20:39:21.0157738Z grep download.pytorch.org /etc/hosts | docker exec -i "${container_name}" sudo bash -c "/bin/cat >> /etc/hosts" 2024-08-06T20:39:21.0158594Z echo "DOCKER_CONTAINER_ID=${container_name}" >> "${GITHUB_ENV}" 2024-08-06T20:39:21.0159406Z docker exec -t "${container_name}" sh -c "pip install $(echo dist/*.whl)[opt-einsum] && ${TEST_COMMAND}" 2024-08-06T20:39:21.0165993Z shell: /usr/bin/bash -e {0} 2024-08-06T20:39:21.0166342Z env: 2024-08-06T20:39:21.0166627Z GIT_DEFAULT_BRANCH: main 2024-08-06T20:39:21.0167057Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:21.0167602Z BUILD_ENVIRONMENT: linux-focal-cuda12.1-py3.10-gcc9 2024-08-06T20:39:21.0168044Z PR_NUMBER: 132710 2024-08-06T20:39:21.0168375Z GITHUB_REPOSITORY: pytorch/pytorch 2024-08-06T20:39:21.0168765Z GITHUB_WORKFLOW: pull 2024-08-06T20:39:21.0169096Z GITHUB_JOB: test 2024-08-06T20:39:21.0169407Z GITHUB_RUN_ID: 10273124344 2024-08-06T20:39:21.0169747Z GITHUB_RUN_NUMBER: 233985 2024-08-06T20:39:21.0170103Z GITHUB_RUN_ATTEMPT: 1 2024-08-06T20:39:21.0170427Z JOB_ID: 28428065757 2024-08-06T20:39:21.0171007Z JOB_NAME: linux-focal-cuda12.1-py3.10-gcc9 / test (default, 1, 5, amz2023.linux.4xlarge.nvidia.gpu) 2024-08-06T20:39:21.0171679Z BRANCH: pull/132710 2024-08-06T20:39:21.0172041Z SHA1: b9d86fa89636e301796d4201f36d86c73f6e49bc 2024-08-06T20:39:21.0172494Z BASE_SHA: 1736af7cf736184c356be1bb00f59fb2feea6d7d 2024-08-06T20:39:21.0172938Z TEST_CONFIG: default 2024-08-06T20:39:21.0173258Z SHARD_NUMBER: 1 2024-08-06T20:39:21.0173544Z NUM_TEST_SHARDS: 5 2024-08-06T20:39:21.0173870Z REENABLED_ISSUES: 2024-08-06T20:39:21.0174201Z CONTINUE_THROUGH_ERROR: False 2024-08-06T20:39:21.0174555Z VERBOSE_TEST_LOGS: False 2024-08-06T20:39:21.0174905Z TEST_SHOWLOCALS: False 2024-08-06T20:39:21.0175243Z NO_TEST_TIMEOUT: False 2024-08-06T20:39:21.0175553Z NO_TD: False 2024-08-06T20:39:21.0175852Z TD_DISTRIBUTED: False 2024-08-06T20:39:21.0176254Z SCCACHE_BUCKET: ossci-compiler-cache-circleci-v2 2024-08-06T20:39:21.0176766Z SCCACHE_S3_KEY_PREFIX: pull 2024-08-06T20:39:21.0177121Z SHM_SIZE: 2g 2024-08-06T20:39:21.0177996Z DOCKER_IMAGE: 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:39:21.0178933Z XLA_CUDA: 2024-08-06T20:39:21.0179391Z XLA_CLANG_CACHE_S3_BUCKET_NAME: ossci-compiler-clang-cache-circleci-xla 2024-08-06T20:39:21.0179974Z PYTORCH_TEST_CUDA_MEM_LEAK_CHECK: 0 2024-08-06T20:39:21.0180368Z PYTORCH_TEST_RERUN_DISABLED_TESTS: 0 2024-08-06T20:39:21.0180767Z DASHBOARD_TAG: 2024-08-06T20:39:21.0181080Z HUGGING_FACE_HUB_TOKEN: 2024-08-06T20:39:21.0181404Z ##[endgroup] 2024-08-06T20:39:21.0208404Z + [[ default == \m\u\l\t\i\g\p\u ]] 2024-08-06T20:39:21.0208876Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *onnx* ]] 2024-08-06T20:39:21.0209342Z + TEST_COMMAND=.ci/pytorch/test.sh 2024-08-06T20:39:21.0218194Z +++ nproc --ignore=2 2024-08-06T20:39:21.0234279Z ++ docker run --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all -e BUILD_ENVIRONMENT -e PR_NUMBER -e GITHUB_ACTIONS -e GITHUB_REPOSITORY -e GITHUB_WORKFLOW -e GITHUB_JOB -e GITHUB_RUN_ID -e GITHUB_RUN_NUMBER -e GITHUB_RUN_ATTEMPT -e JOB_ID -e JOB_NAME -e BASE_SHA -e BRANCH -e SHA1 -e AWS_DEFAULT_REGION -e IN_WHEEL_TEST -e SHARD_NUMBER -e TEST_CONFIG -e NUM_TEST_SHARDS -e REENABLED_ISSUES -e CONTINUE_THROUGH_ERROR -e VERBOSE_TEST_LOGS -e TEST_SHOWLOCALS -e NO_TEST_TIMEOUT -e NO_TD -e TD_DISTRIBUTED -e PR_LABELS -e MAX_JOBS=14 -e SCCACHE_BUCKET -e SCCACHE_S3_KEY_PREFIX -e XLA_CUDA -e XLA_CLANG_CACHE_S3_BUCKET_NAME -e PYTORCH_TEST_CUDA_MEM_LEAK_CHECK -e PYTORCH_TEST_RERUN_DISABLED_TESTS -e SKIP_SCCACHE_INITIALIZATION=1 -e HUGGING_FACE_HUB_TOKEN -e DASHBOARD_TAG --env-file=/tmp/github_env_10273124344 --security-opt seccomp=unconfined --cap-add=SYS_PTRACE --ipc=host --shm-size=2g --tty --detach --name= --user jenkins -v /home/ec2-user/actions-runner/_work/pytorch/pytorch:/var/lib/jenkins/workspace -w /var/lib/jenkins/workspace 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T20:39:32.0944287Z + container_name=492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T20:39:32.0946793Z + grep download.pytorch.org /etc/hosts 2024-08-06T20:39:32.0948363Z + docker exec -i 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e sudo bash -c '/bin/cat >> /etc/hosts' 2024-08-06T20:39:32.2749983Z + echo DOCKER_CONTAINER_ID=492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T20:39:32.2754967Z ++ echo dist/torch-2.5.0a0+gitb9d86fa-cp310-cp310-linux_x86_64.whl 2024-08-06T20:39:32.2757774Z + docker exec -t 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e sh -c 'pip install dist/torch-2.5.0a0+gitb9d86fa-cp310-cp310-linux_x86_64.whl[opt-einsum] && .ci/pytorch/test.sh' 2024-08-06T20:39:32.8136567Z Processing ./dist/torch-2.5.0a0+gitb9d86fa-cp310-cp310-linux_x86_64.whl (from torch==2.5.0a0+gitb9d86fa) 2024-08-06T20:39:33.9199916Z Requirement already satisfied: filelock in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torch==2.5.0a0+gitb9d86fa->torch==2.5.0a0+gitb9d86fa) (3.13.1) 2024-08-06T20:39:33.9206645Z Requirement already satisfied: typing-extensions>=4.8.0 in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torch==2.5.0a0+gitb9d86fa->torch==2.5.0a0+gitb9d86fa) (4.12.2) 2024-08-06T20:39:33.9211358Z Requirement already satisfied: networkx in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torch==2.5.0a0+gitb9d86fa->torch==2.5.0a0+gitb9d86fa) (2.8.8) 2024-08-06T20:39:33.9215924Z Requirement already satisfied: jinja2 in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torch==2.5.0a0+gitb9d86fa->torch==2.5.0a0+gitb9d86fa) (3.1.4) 2024-08-06T20:39:33.9220314Z Requirement already satisfied: fsspec in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torch==2.5.0a0+gitb9d86fa->torch==2.5.0a0+gitb9d86fa) (2024.6.1) 2024-08-06T20:39:33.9233439Z Requirement already satisfied: sympy>=1.13.0 in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torch==2.5.0a0+gitb9d86fa->torch==2.5.0a0+gitb9d86fa) (1.13.1) 2024-08-06T20:39:33.9273177Z Requirement already satisfied: opt-einsum>=3.3 in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torch==2.5.0a0+gitb9d86fa->torch==2.5.0a0+gitb9d86fa) (3.3.0) 2024-08-06T20:39:33.9352822Z Requirement already satisfied: numpy>=1.7 in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from opt-einsum>=3.3->torch==2.5.0a0+gitb9d86fa->torch==2.5.0a0+gitb9d86fa) (1.21.2) 2024-08-06T20:39:33.9398994Z Requirement already satisfied: mpmath<1.4,>=1.1.0 in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from sympy>=1.13.0->torch==2.5.0a0+gitb9d86fa->torch==2.5.0a0+gitb9d86fa) (1.3.0) 2024-08-06T20:39:34.0690206Z Requirement already satisfied: MarkupSafe>=2.0 in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from jinja2->torch==2.5.0a0+gitb9d86fa->torch==2.5.0a0+gitb9d86fa) (2.1.5) 2024-08-06T20:39:35.2066653Z Installing collected packages: torch 2024-08-06T20:39:48.3810671Z Successfully installed torch-2.5.0a0+gitb9d86fa 2024-08-06T20:39:48.4499135Z ++ dirname .ci/pytorch/test.sh 2024-08-06T20:39:48.4507258Z + source .ci/pytorch/common.sh 2024-08-06T20:39:48.4510656Z +++ dirname .ci/pytorch/common.sh 2024-08-06T20:39:48.4518935Z ++ source .ci/pytorch/common_utils.sh 2024-08-06T20:39:48.4521690Z +++ declare -f -t trap_add 2024-08-06T20:39:48.4528363Z ++ set -ex 2024-08-06T20:39:48.4529014Z ++ [[ linux-focal-cuda12.1-py3.10-gcc9 == *rocm* ]] 2024-08-06T20:39:48.4529895Z ++ BUILD_TEST_LIBTORCH=0 2024-08-06T20:39:48.4530731Z + [[ linux-focal-cuda12.1-py3.10-gcc9 != *rocm* ]] 2024-08-06T20:39:48.4534031Z ++ stat -c %u /var/lib/jenkins/workspace 2024-08-06T20:39:48.4550760Z + WORKSPACE_ORIGINAL_OWNER_ID=1000 2024-08-06T20:39:48.4551457Z + trap_add cleanup_workspace EXIT 2024-08-06T20:39:48.4551875Z + trap_add_cmd=cleanup_workspace 2024-08-06T20:39:48.4552247Z + shift 2024-08-06T20:39:48.4552671Z + for trap_add_name in "$@" 2024-08-06T20:39:48.4560137Z +++ trap -p EXIT 2024-08-06T20:39:48.4563530Z ++ eval 'extract_trap_cmd ' 2024-08-06T20:39:48.4563929Z +++ extract_trap_cmd 2024-08-06T20:39:48.4564270Z +++ printf '%s\n' '' 2024-08-06T20:39:48.4564687Z ++ printf '%s\n' cleanup_workspace 2024-08-06T20:39:48.4567057Z + trap -- ' 2024-08-06T20:39:48.4567375Z cleanup_workspace' EXIT 2024-08-06T20:39:48.4567801Z + sudo chown -R jenkins /var/lib/jenkins/workspace 2024-08-06T20:39:49.0488734Z + git config --global --add safe.directory /var/lib/jenkins/workspace 2024-08-06T20:39:49.0509105Z + echo 'Environment variables:' 2024-08-06T20:39:49.0509515Z Environment variables: 2024-08-06T20:39:49.0509839Z + env 2024-08-06T20:39:49.0519462Z INSTALLED_DB=yes 2024-08-06T20:39:49.0520004Z NV_LIBCUBLAS_VERSION=12.1.3.1-1 2024-08-06T20:39:49.0520405Z NVIDIA_VISIBLE_DEVICES=all 2024-08-06T20:39:49.0520756Z NV_NVML_DEV_VERSION=12.1.105-1 2024-08-06T20:39:49.0521291Z GITHUB_WORKSPACE=/home/ec2-user/actions-runner/_work/pytorch/pytorch 2024-08-06T20:39:49.0521854Z CONTINUE_THROUGH_ERROR=False 2024-08-06T20:39:49.0522264Z NV_LIBNCCL_DEV_PACKAGE=libnccl-dev=2.17.1-1+cuda12.1 2024-08-06T20:39:49.0522741Z NV_LIBNCCL_DEV_PACKAGE_VERSION=2.17.1-1 2024-08-06T20:39:49.0523218Z BUILD_ENVIRONMENT=linux-focal-cuda12.1-py3.10-gcc9 2024-08-06T20:39:49.0523657Z HOSTNAME=492035230bda 2024-08-06T20:39:49.0524369Z GITHUB_PATH=/home/ec2-user/actions-runner/_work/_temp/_runner_file_commands/add_path_2748003d-9015-44a8-ba83-11a0dd1e1a38 2024-08-06T20:39:49.0525166Z GITHUB_ACTION=__self 2024-08-06T20:39:49.0525498Z PYTORCH_TEST_CUDA_MEM_LEAK_CHECK=0 2024-08-06T20:39:49.0529440Z NVIDIA_REQUIRE_CUDA=cuda>=12.1 brand=tesla,driver>=470,driver<471 brand=unknown,driver>=470,driver<471 brand=nvidia,driver>=470,driver<471 brand=nvidiartx,driver>=470,driver<471 brand=geforce,driver>=470,driver<471 brand=geforcertx,driver>=470,driver<471 brand=quadro,driver>=470,driver<471 brand=quadrortx,driver>=470,driver<471 brand=titan,driver>=470,driver<471 brand=titanrtx,driver>=470,driver<471 brand=tesla,driver>=525,driver<526 brand=unknown,driver>=525,driver<526 brand=nvidia,driver>=525,driver<526 brand=nvidiartx,driver>=525,driver<526 brand=geforce,driver>=525,driver<526 brand=geforcertx,driver>=525,driver<526 brand=quadro,driver>=525,driver<526 brand=quadrortx,driver>=525,driver<526 brand=titan,driver>=525,driver<526 brand=titanrtx,driver>=525,driver<526 2024-08-06T20:39:49.0533115Z NV_LIBCUBLAS_DEV_PACKAGE=libcublas-dev-12-1=12.1.3.1-1 2024-08-06T20:39:49.0533592Z NV_NVTX_VERSION=12.1.105-1 2024-08-06T20:39:49.0533955Z GITHUB_RUN_NUMBER=233985 2024-08-06T20:39:49.0534291Z TEST_CONFIG=default 2024-08-06T20:39:49.0534634Z GITHUB_REPOSITORY_OWNER_ID=21003710 2024-08-06T20:39:49.0535059Z TORCH_NVCC_FLAGS=-Xfatbin -compress-all 2024-08-06T20:39:49.0535469Z NV_CUDA_CUDART_DEV_VERSION=12.1.105-1 2024-08-06T20:39:49.0535882Z NV_LIBCUSPARSE_VERSION=12.1.0.106-1 2024-08-06T20:39:49.0536279Z NV_LIBNPP_VERSION=12.1.0.40-1 2024-08-06T20:39:49.0536643Z GITHUB_TRIGGERING_ACTOR=drisspg 2024-08-06T20:39:49.0537080Z CMAKE_CUDA_COMPILER_LAUNCHER=/opt/cache/bin/sccache 2024-08-06T20:39:49.0537538Z GITHUB_REF_TYPE=branch 2024-08-06T20:39:49.0537866Z TORCH_CUDA_ARCH_LIST=Maxwell 2024-08-06T20:39:49.0538227Z NCCL_VERSION=2.17.1-1 2024-08-06T20:39:49.0538609Z BASE_SHA=1736af7cf736184c356be1bb00f59fb2feea6d7d 2024-08-06T20:39:49.0539019Z XLA_CUDA= 2024-08-06T20:39:49.0539312Z HUGGING_FACE_HUB_TOKEN= 2024-08-06T20:39:49.0540222Z *** 2024-08-06T20:39:49.0540500Z CARGO_NET_GIT_FETCH_WITH_CLI=true 2024-08-06T20:39:49.0540907Z GITHUB_REPOSITORY_ID=65600975 2024-08-06T20:39:49.0541273Z GITHUB_ACTIONS=true 2024-08-06T20:39:49.0541738Z NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:49.0542184Z NV_NVPROF_DEV_PACKAGE=cuda-nvprof-12-1=12.1.105-1 2024-08-06T20:39:49.0542656Z NV_LIBNPP_PACKAGE=libnpp-12-1=12.1.0.40-1 2024-08-06T20:39:49.0543080Z SHA1=b9d86fa89636e301796d4201f36d86c73f6e49bc 2024-08-06T20:39:49.0543535Z NV_LIBNCCL_DEV_PACKAGE_NAME=libnccl-dev 2024-08-06T20:39:49.0543992Z GITHUB_SHA=bf5bb5a1585a03379137fab341e87c02c77e76cd 2024-08-06T20:39:49.0544644Z GITHUB_WORKFLOW_REF=pytorch/pytorch/.github/workflows/pull.yml@refs/pull/132710/merge 2024-08-06T20:39:49.0545258Z UCC_HOME=/usr 2024-08-06T20:39:49.0545572Z NV_LIBCUBLAS_DEV_VERSION=12.1.3.1-1 2024-08-06T20:39:49.0545942Z VERBOSE_TEST_LOGS=False 2024-08-06T20:39:49.0546280Z NVIDIA_PRODUCT_NAME=CUDA 2024-08-06T20:39:49.0546664Z NV_LIBCUBLAS_DEV_PACKAGE_NAME=libcublas-dev-12-1 2024-08-06T20:39:49.0547125Z GITHUB_REF=refs/pull/132710/merge 2024-08-06T20:39:49.0547512Z NV_CUDA_CUDART_VERSION=12.1.105-1 2024-08-06T20:39:49.0547861Z SHARD_NUMBER=1 2024-08-06T20:39:49.0548183Z GITHUB_REF_PROTECTED=false 2024-08-06T20:39:49.0548539Z HOME=/var/lib/jenkins 2024-08-06T20:39:49.0548884Z GITHUB_API_URL=https://api.github.com 2024-08-06T20:39:49.0549312Z PYTORCH_TEST_RERUN_DISABLED_TESTS=0 2024-08-06T20:39:49.0549760Z UCX_COMMIT=7bb2722ff2187a0cad557ae4a6afa090569f83fb 2024-08-06T20:39:49.0550200Z SCCACHE_S3_KEY_PREFIX=pull 2024-08-06T20:39:49.0550549Z CUDA_VERSION=12.1.1 2024-08-06T20:39:49.0550909Z NV_LIBCUBLAS_PACKAGE=libcublas-12-1=12.1.3.1-1 2024-08-06T20:39:49.0551315Z NUM_TEST_SHARDS=5 2024-08-06T20:39:49.0551614Z UCX_HOME=/usr 2024-08-06T20:39:49.0552052Z NV_CUDA_NSIGHT_COMPUTE_DEV_PACKAGE=cuda-nsight-compute-12-1=12.1.1-1 2024-08-06T20:39:49.0552961Z GITHUB_STATE=/home/ec2-user/actions-runner/_work/_temp/_runner_file_commands/save_state_2748003d-9015-44a8-ba83-11a0dd1e1a38 2024-08-06T20:39:49.0554050Z JOB_NAME=linux-focal-cuda12.1-py3.10-gcc9 / test (default, 1, 5, amz2023.linux.4xlarge.nvidia.gpu) 2024-08-06T20:39:49.0555086Z GITHUB_ENV=/home/ec2-user/actions-runner/_work/_temp/_runner_file_commands/set_env_2748003d-9015-44a8-ba83-11a0dd1e1a38 2024-08-06T20:39:49.0556054Z GITHUB_EVENT_PATH=/home/ec2-user/actions-runner/_work/_temp/_github_workflow/event.json 2024-08-06T20:39:49.0556692Z GITHUB_EVENT_NAME=pull_request 2024-08-06T20:39:49.0557062Z DASHBOARD_TAG= 2024-08-06T20:39:49.0557354Z GITHUB_RUN_ID=10273124344 2024-08-06T20:39:49.0557757Z NV_LIBNPP_DEV_PACKAGE=libnpp-dev-12-1=12.1.0.40-1 2024-08-06T20:39:49.0558238Z NV_LIBCUBLAS_PACKAGE_NAME=libcublas-12-1 2024-08-06T20:39:49.0559068Z GITHUB_STEP_SUMMARY=/home/ec2-user/actions-runner/_work/_temp/_runner_file_commands/step_summary_2748003d-9015-44a8-ba83-11a0dd1e1a38 2024-08-06T20:39:49.0559890Z GITHUB_ACTOR=drisspg 2024-08-06T20:39:49.0560234Z NV_LIBNPP_DEV_VERSION=12.1.0.40-1 2024-08-06T20:39:49.0560601Z PR_NUMBER=132710 2024-08-06T20:39:49.0560918Z GITHUB_RUN_ATTEMPT=1 2024-08-06T20:39:49.0561260Z ANACONDA_PYTHON_VERSION=3.10 2024-08-06T20:39:49.0561677Z GITHUB_GRAPHQL_URL=https://api.github.com/graphql 2024-08-06T20:39:49.0562132Z TERM=xterm 2024-08-06T20:39:49.0562453Z NV_LIBCUSPARSE_DEV_VERSION=12.1.0.106-1 2024-08-06T20:39:49.0562840Z INSTALLED_VISION=yes 2024-08-06T20:39:49.0563168Z BRANCH=pull/132710 2024-08-06T20:39:49.0563502Z OPENSSL_ROOT_DIR=/opt/openssl 2024-08-06T20:39:49.0563881Z LIBRARY_PATH=/usr/local/cuda/lib64/stubs 2024-08-06T20:39:49.0564293Z CUDA_PATH=/usr/local/cuda 2024-08-06T20:39:49.0564953Z GITHUB_ACTION_PATH=/home/ec2-user/actions-runner/_work/pytorch/pytorch/./.github/actions/setup-linux 2024-08-06T20:39:49.0565655Z GITHUB_SERVER_URL=https://github.com 2024-08-06T20:39:49.0566120Z UCC_COMMIT=20eae37090a4ce1b32bcce6144ccad0b49943e0b 2024-08-06T20:39:49.0566565Z REENABLED_ISSUES= 2024-08-06T20:39:49.0566847Z SHLVL=1 2024-08-06T20:39:49.0567116Z MAX_JOBS=14 2024-08-06T20:39:49.0567512Z NV_CUDA_LIB_VERSION=12.1.1-1 2024-08-06T20:39:49.0567844Z NVARCH=x86_64 2024-08-06T20:39:49.0568143Z GITHUB_ACTOR_ID=32754868 2024-08-06T20:39:49.0568663Z GITHUB_WORKFLOW_SHA=bf5bb5a1585a03379137fab341e87c02c77e76cd 2024-08-06T20:39:49.0569150Z GITHUB_REF_NAME=132710/merge 2024-08-06T20:39:49.0569534Z NV_CUDA_COMPAT_PACKAGE=cuda-compat-12-1 2024-08-06T20:39:49.0570101Z XLA_CLANG_CACHE_S3_BUCKET_NAME=ossci-compiler-clang-cache-circleci-xla 2024-08-06T20:39:49.0570628Z GITHUB_JOB=test 2024-08-06T20:39:49.0570971Z NV_LIBNCCL_PACKAGE=libnccl2=2.17.1-1+cuda12.1 2024-08-06T20:39:49.0571494Z LD_LIBRARY_PATH=/usr/local/nvidia/lib:/usr/local/nvidia/lib64 2024-08-06T20:39:49.0571968Z NO_TEST_TIMEOUT=False 2024-08-06T20:39:49.0572296Z TD_DISTRIBUTED=False 2024-08-06T20:39:49.0572628Z NV_CUDA_NSIGHT_COMPUTE_VERSION=12.1.1-1 2024-08-06T20:39:49.0573047Z GITHUB_REPOSITORY=pytorch/pytorch 2024-08-06T20:39:49.0573439Z NV_NVPROF_VERSION=12.1.105-1 2024-08-06T20:39:49.0573786Z GITHUB_RETENTION_DAYS=90 2024-08-06T20:39:49.0574142Z OPENSSL_DIR=/opt/openssl 2024-08-06T20:39:49.0574494Z GITHUB_ACTION_REPOSITORY= 2024-08-06T20:39:49.0575432Z PATH=/opt/cache/bin:/opt/conda/envs/py_3.10/bin:/opt/conda/bin:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin 2024-08-06T20:39:49.0576439Z GITHUB_BASE_REF=gh/drisspg/23/base 2024-08-06T20:39:49.0576847Z NV_LIBNCCL_PACKAGE_NAME=libnccl2 2024-08-06T20:39:49.0577197Z CI=true 2024-08-06T20:39:49.0577496Z NV_LIBNCCL_PACKAGE_VERSION=2.17.1-1 2024-08-06T20:39:49.0577900Z GITHUB_REPOSITORY_OWNER=pytorch 2024-08-06T20:39:49.0578252Z JOB_ID=28428065757 2024-08-06T20:39:49.0578570Z INSTALLED_PROTOBUF=yes 2024-08-06T20:39:49.0578926Z GITHUB_HEAD_REF=gh/drisspg/23/head 2024-08-06T20:39:49.0579347Z GITHUB_ACTION_REF= 2024-08-06T20:39:49.0579732Z SCCACHE_BUCKET=ossci-compiler-cache-circleci-v2 2024-08-06T20:39:49.0580188Z TEST_SHOWLOCALS=False 2024-08-06T20:39:49.0580509Z GITHUB_WORKFLOW=pull 2024-08-06T20:39:49.0580857Z DEBIAN_FRONTEND=noninteractive 2024-08-06T20:39:49.0581637Z GITHUB_OUTPUT=/home/ec2-user/actions-runner/_work/_temp/_runner_file_commands/set_output_2748003d-9015-44a8-ba83-11a0dd1e1a38 2024-08-06T20:39:49.0582404Z NO_TD=False 2024-08-06T20:39:49.0582723Z SKIP_SCCACHE_INITIALIZATION=1 2024-08-06T20:39:49.0583088Z _=/usr/bin/env 2024-08-06T20:39:49.0583494Z ++ python -c 'import site; print(site.getsitepackages()[0])' 2024-08-06T20:39:49.0744054Z + TORCH_INSTALL_DIR=/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch 2024-08-06T20:39:49.0744851Z + TORCH_BIN_DIR=/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/bin 2024-08-06T20:39:49.0745587Z + TORCH_LIB_DIR=/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib 2024-08-06T20:39:49.0746315Z + TORCH_TEST_DIR=/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/test 2024-08-06T20:39:49.0747187Z + BUILD_DIR=build 2024-08-06T20:39:49.0747763Z + BUILD_RENAMED_DIR=build_renamed 2024-08-06T20:39:49.0748154Z + BUILD_BIN_DIR=build/bin 2024-08-06T20:39:49.0748496Z + SHARD_NUMBER=1 2024-08-06T20:39:49.0748818Z + NUM_TEST_SHARDS=5 2024-08-06T20:39:49.0749118Z + export VALGRIND=ON 2024-08-06T20:39:49.0749437Z + VALGRIND=ON 2024-08-06T20:39:49.0749819Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *clang9* ]] 2024-08-06T20:39:49.0750253Z + [[ 0 == \1 ]] 2024-08-06T20:39:49.0750547Z + [[ False == \1 ]] 2024-08-06T20:39:49.0750923Z + [[ linux-focal-cuda12.1-py3.10-gcc9 != *bazel* ]] 2024-08-06T20:39:49.0754385Z ++ realpath build/custom_test_artifacts 2024-08-06T20:39:49.0764625Z + CUSTOM_TEST_ARTIFACT_BUILD_DIR=/var/lib/jenkins/workspace/build/custom_test_artifacts 2024-08-06T20:39:49.0765261Z + [[ -n '' ]] 2024-08-06T20:39:49.0765561Z + echo 'Environment variables' 2024-08-06T20:39:49.0765934Z Environment variables 2024-08-06T20:39:49.0766252Z + env 2024-08-06T20:39:49.0773377Z INSTALLED_DB=yes 2024-08-06T20:39:49.0773999Z NV_LIBCUBLAS_VERSION=12.1.3.1-1 2024-08-06T20:39:49.0774731Z NVIDIA_VISIBLE_DEVICES=all 2024-08-06T20:39:49.0775401Z NV_NVML_DEV_VERSION=12.1.105-1 2024-08-06T20:39:49.0776555Z GITHUB_WORKSPACE=/home/ec2-user/actions-runner/_work/pytorch/pytorch 2024-08-06T20:39:49.0777112Z CONTINUE_THROUGH_ERROR=False 2024-08-06T20:39:49.0777665Z NV_LIBNCCL_DEV_PACKAGE=libnccl-dev=2.17.1-1+cuda12.1 2024-08-06T20:39:49.0778164Z NV_LIBNCCL_DEV_PACKAGE_VERSION=2.17.1-1 2024-08-06T20:39:49.0778685Z BUILD_ENVIRONMENT=linux-focal-cuda12.1-py3.10-gcc9 2024-08-06T20:39:49.0779319Z HOSTNAME=492035230bda 2024-08-06T20:39:49.0780028Z GITHUB_PATH=/home/ec2-user/actions-runner/_work/_temp/_runner_file_commands/add_path_2748003d-9015-44a8-ba83-11a0dd1e1a38 2024-08-06T20:39:49.0780795Z GITHUB_ACTION=__self 2024-08-06T20:39:49.0781121Z PYTORCH_TEST_CUDA_MEM_LEAK_CHECK=0 2024-08-06T20:39:49.0785087Z NVIDIA_REQUIRE_CUDA=cuda>=12.1 brand=tesla,driver>=470,driver<471 brand=unknown,driver>=470,driver<471 brand=nvidia,driver>=470,driver<471 brand=nvidiartx,driver>=470,driver<471 brand=geforce,driver>=470,driver<471 brand=geforcertx,driver>=470,driver<471 brand=quadro,driver>=470,driver<471 brand=quadrortx,driver>=470,driver<471 brand=titan,driver>=470,driver<471 brand=titanrtx,driver>=470,driver<471 brand=tesla,driver>=525,driver<526 brand=unknown,driver>=525,driver<526 brand=nvidia,driver>=525,driver<526 brand=nvidiartx,driver>=525,driver<526 brand=geforce,driver>=525,driver<526 brand=geforcertx,driver>=525,driver<526 brand=quadro,driver>=525,driver<526 brand=quadrortx,driver>=525,driver<526 brand=titan,driver>=525,driver<526 brand=titanrtx,driver>=525,driver<526 2024-08-06T20:39:49.0789585Z NV_LIBCUBLAS_DEV_PACKAGE=libcublas-dev-12-1=12.1.3.1-1 2024-08-06T20:39:49.0790042Z NV_NVTX_VERSION=12.1.105-1 2024-08-06T20:39:49.0790397Z GITHUB_RUN_NUMBER=233985 2024-08-06T20:39:49.0790719Z TEST_CONFIG=default 2024-08-06T20:39:49.0791052Z GITHUB_REPOSITORY_OWNER_ID=21003710 2024-08-06T20:39:49.0791476Z TORCH_NVCC_FLAGS=-Xfatbin -compress-all 2024-08-06T20:39:49.0791881Z NV_CUDA_CUDART_DEV_VERSION=12.1.105-1 2024-08-06T20:39:49.0792289Z NV_LIBCUSPARSE_VERSION=12.1.0.106-1 2024-08-06T20:39:49.0792685Z NV_LIBNPP_VERSION=12.1.0.40-1 2024-08-06T20:39:49.0793058Z GITHUB_TRIGGERING_ACTOR=drisspg 2024-08-06T20:39:49.0793490Z CMAKE_CUDA_COMPILER_LAUNCHER=/opt/cache/bin/sccache 2024-08-06T20:39:49.0793943Z GITHUB_REF_TYPE=branch 2024-08-06T20:39:49.0794279Z TORCH_CUDA_ARCH_LIST=Maxwell 2024-08-06T20:39:49.0794642Z NCCL_VERSION=2.17.1-1 2024-08-06T20:39:49.0795023Z BASE_SHA=1736af7cf736184c356be1bb00f59fb2feea6d7d 2024-08-06T20:39:49.0795432Z XLA_CUDA= 2024-08-06T20:39:49.0795724Z HUGGING_FACE_HUB_TOKEN= 2024-08-06T20:39:49.0796146Z *** 2024-08-06T20:39:49.0796422Z CARGO_NET_GIT_FETCH_WITH_CLI=true 2024-08-06T20:39:49.0796822Z GITHUB_REPOSITORY_ID=65600975 2024-08-06T20:39:49.0797188Z GITHUB_ACTIONS=true 2024-08-06T20:39:49.0797505Z NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T20:39:49.0797929Z NV_NVPROF_DEV_PACKAGE=cuda-nvprof-12-1=12.1.105-1 2024-08-06T20:39:49.0798404Z NV_LIBNPP_PACKAGE=libnpp-12-1=12.1.0.40-1 2024-08-06T20:39:49.0798834Z SHA1=b9d86fa89636e301796d4201f36d86c73f6e49bc 2024-08-06T20:39:49.0799276Z NV_LIBNCCL_DEV_PACKAGE_NAME=libnccl-dev 2024-08-06T20:39:49.0799748Z GITHUB_SHA=bf5bb5a1585a03379137fab341e87c02c77e76cd 2024-08-06T20:39:49.0800450Z GITHUB_WORKFLOW_REF=pytorch/pytorch/.github/workflows/pull.yml@refs/pull/132710/merge 2024-08-06T20:39:49.0801053Z UCC_HOME=/usr 2024-08-06T20:39:49.0801369Z NV_LIBCUBLAS_DEV_VERSION=12.1.3.1-1 2024-08-06T20:39:49.0801762Z VERBOSE_TEST_LOGS=False 2024-08-06T20:39:49.0802086Z NVIDIA_PRODUCT_NAME=CUDA 2024-08-06T20:39:49.0802496Z NV_LIBCUBLAS_DEV_PACKAGE_NAME=libcublas-dev-12-1 2024-08-06T20:39:49.0802951Z GITHUB_REF=refs/pull/132710/merge 2024-08-06T20:39:49.0803326Z NV_CUDA_CUDART_VERSION=12.1.105-1 2024-08-06T20:39:49.0803700Z SHARD_NUMBER=1 2024-08-06T20:39:49.0804478Z GITHUB_REF_PROTECTED=false 2024-08-06T20:39:49.0804836Z HOME=/var/lib/jenkins 2024-08-06T20:39:49.0805202Z GITHUB_API_URL=https://api.github.com 2024-08-06T20:39:49.0805611Z PYTORCH_TEST_RERUN_DISABLED_TESTS=0 2024-08-06T20:39:49.0806059Z UCX_COMMIT=7bb2722ff2187a0cad557ae4a6afa090569f83fb 2024-08-06T20:39:49.0806708Z SCCACHE_S3_KEY_PREFIX=pull 2024-08-06T20:39:49.0807042Z CUDA_VERSION=12.1.1 2024-08-06T20:39:49.0807400Z NV_LIBCUBLAS_PACKAGE=libcublas-12-1=12.1.3.1-1 2024-08-06T20:39:49.0807836Z NUM_TEST_SHARDS=5 2024-08-06T20:39:49.0808258Z UCX_HOME=/usr 2024-08-06T20:39:49.0808719Z NV_CUDA_NSIGHT_COMPUTE_DEV_PACKAGE=cuda-nsight-compute-12-1=12.1.1-1 2024-08-06T20:39:49.0809659Z GITHUB_STATE=/home/ec2-user/actions-runner/_work/_temp/_runner_file_commands/save_state_2748003d-9015-44a8-ba83-11a0dd1e1a38 2024-08-06T20:39:49.0810693Z JOB_NAME=linux-focal-cuda12.1-py3.10-gcc9 / test (default, 1, 5, amz2023.linux.4xlarge.nvidia.gpu) 2024-08-06T20:39:49.0811716Z GITHUB_ENV=/home/ec2-user/actions-runner/_work/_temp/_runner_file_commands/set_env_2748003d-9015-44a8-ba83-11a0dd1e1a38 2024-08-06T20:39:49.0812693Z GITHUB_EVENT_PATH=/home/ec2-user/actions-runner/_work/_temp/_github_workflow/event.json 2024-08-06T20:39:49.0813327Z GITHUB_EVENT_NAME=pull_request 2024-08-06T20:39:49.0813677Z DASHBOARD_TAG= 2024-08-06T20:39:49.0813984Z GITHUB_RUN_ID=10273124344 2024-08-06T20:39:49.0814393Z NV_LIBNPP_DEV_PACKAGE=libnpp-dev-12-1=12.1.0.40-1 2024-08-06T20:39:49.0814853Z NV_LIBCUBLAS_PACKAGE_NAME=libcublas-12-1 2024-08-06T20:39:49.0815708Z GITHUB_STEP_SUMMARY=/home/ec2-user/actions-runner/_work/_temp/_runner_file_commands/step_summary_2748003d-9015-44a8-ba83-11a0dd1e1a38 2024-08-06T20:39:49.0816531Z GITHUB_ACTOR=drisspg 2024-08-06T20:39:49.0816856Z NV_LIBNPP_DEV_VERSION=12.1.0.40-1 2024-08-06T20:39:49.0817234Z PR_NUMBER=132710 2024-08-06T20:39:49.0817543Z GITHUB_RUN_ATTEMPT=1 2024-08-06T20:39:49.0817845Z VALGRIND=ON 2024-08-06T20:39:49.0818150Z ANACONDA_PYTHON_VERSION=3.10 2024-08-06T20:39:49.0818566Z GITHUB_GRAPHQL_URL=https://api.github.com/graphql 2024-08-06T20:39:49.0819013Z TERM=xterm 2024-08-06T20:39:49.0819331Z NV_LIBCUSPARSE_DEV_VERSION=12.1.0.106-1 2024-08-06T20:39:49.0819718Z INSTALLED_VISION=yes 2024-08-06T20:39:49.0820048Z BRANCH=pull/132710 2024-08-06T20:39:49.0820379Z OPENSSL_ROOT_DIR=/opt/openssl 2024-08-06T20:39:49.0820753Z LIBRARY_PATH=/usr/local/cuda/lib64/stubs 2024-08-06T20:39:49.0821177Z CUDA_PATH=/usr/local/cuda 2024-08-06T20:39:49.0821836Z GITHUB_ACTION_PATH=/home/ec2-user/actions-runner/_work/pytorch/pytorch/./.github/actions/setup-linux 2024-08-06T20:39:49.0822540Z GITHUB_SERVER_URL=https://github.com 2024-08-06T20:39:49.0823003Z UCC_COMMIT=20eae37090a4ce1b32bcce6144ccad0b49943e0b 2024-08-06T20:39:49.0823456Z REENABLED_ISSUES= 2024-08-06T20:39:49.0823743Z SHLVL=1 2024-08-06T20:39:49.0824010Z MAX_JOBS=14 2024-08-06T20:39:49.0824311Z NV_CUDA_LIB_VERSION=12.1.1-1 2024-08-06T20:39:49.0824646Z NVARCH=x86_64 2024-08-06T20:39:49.0824946Z GITHUB_ACTOR_ID=32754868 2024-08-06T20:39:49.0825388Z GITHUB_WORKFLOW_SHA=bf5bb5a1585a03379137fab341e87c02c77e76cd 2024-08-06T20:39:49.0825869Z GITHUB_REF_NAME=132710/merge 2024-08-06T20:39:49.0826254Z NV_CUDA_COMPAT_PACKAGE=cuda-compat-12-1 2024-08-06T20:39:49.0826829Z XLA_CLANG_CACHE_S3_BUCKET_NAME=ossci-compiler-clang-cache-circleci-xla 2024-08-06T20:39:49.0827355Z GITHUB_JOB=test 2024-08-06T20:39:49.0827705Z NV_LIBNCCL_PACKAGE=libnccl2=2.17.1-1+cuda12.1 2024-08-06T20:39:49.0828223Z LD_LIBRARY_PATH=/usr/local/nvidia/lib:/usr/local/nvidia/lib64 2024-08-06T20:39:49.0828694Z NO_TEST_TIMEOUT=False 2024-08-06T20:39:49.0829028Z TD_DISTRIBUTED=False 2024-08-06T20:39:49.0829377Z NV_CUDA_NSIGHT_COMPUTE_VERSION=12.1.1-1 2024-08-06T20:39:49.0829780Z GITHUB_REPOSITORY=pytorch/pytorch 2024-08-06T20:39:49.0830173Z NV_NVPROF_VERSION=12.1.105-1 2024-08-06T20:39:49.0830538Z GITHUB_RETENTION_DAYS=90 2024-08-06T20:39:49.0830869Z OPENSSL_DIR=/opt/openssl 2024-08-06T20:39:49.0831221Z GITHUB_ACTION_REPOSITORY= 2024-08-06T20:39:49.0832145Z PATH=/opt/cache/bin:/opt/conda/envs/py_3.10/bin:/opt/conda/bin:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin 2024-08-06T20:39:49.0833135Z GITHUB_BASE_REF=gh/drisspg/23/base 2024-08-06T20:39:49.0833539Z NV_LIBNCCL_PACKAGE_NAME=libnccl2 2024-08-06T20:39:49.0833901Z CI=true 2024-08-06T20:39:49.0834177Z NV_LIBNCCL_PACKAGE_VERSION=2.17.1-1 2024-08-06T20:39:49.0834684Z GITHUB_REPOSITORY_OWNER=pytorch 2024-08-06T20:39:49.0835032Z JOB_ID=28428065757 2024-08-06T20:39:49.0835349Z INSTALLED_PROTOBUF=yes 2024-08-06T20:39:49.0835778Z GITHUB_HEAD_REF=gh/drisspg/23/head 2024-08-06T20:39:49.0836155Z GITHUB_ACTION_REF= 2024-08-06T20:39:49.0836535Z SCCACHE_BUCKET=ossci-compiler-cache-circleci-v2 2024-08-06T20:39:49.0836986Z TEST_SHOWLOCALS=False 2024-08-06T20:39:49.0837299Z GITHUB_WORKFLOW=pull 2024-08-06T20:39:49.0837645Z DEBIAN_FRONTEND=noninteractive 2024-08-06T20:39:49.0838420Z GITHUB_OUTPUT=/home/ec2-user/actions-runner/_work/_temp/_runner_file_commands/set_output_2748003d-9015-44a8-ba83-11a0dd1e1a38 2024-08-06T20:39:49.0839181Z NO_TD=False 2024-08-06T20:39:49.0839490Z SKIP_SCCACHE_INITIALIZATION=1 2024-08-06T20:39:49.0839851Z _=/usr/bin/env 2024-08-06T20:39:49.0840143Z + echo 'Testing pytorch' 2024-08-06T20:39:49.0840485Z Testing pytorch 2024-08-06T20:39:49.0840819Z + export LANG=C.UTF-8 2024-08-06T20:39:49.0841125Z + LANG=C.UTF-8 2024-08-06T20:39:49.0841437Z + PR_NUMBER=132710 2024-08-06T20:39:49.0841758Z + [[ default == \d\e\f\a\u\l\t ]] 2024-08-06T20:39:49.0842129Z + export CUDA_VISIBLE_DEVICES=0 2024-08-06T20:39:49.0842502Z + CUDA_VISIBLE_DEVICES=0 2024-08-06T20:39:49.0842860Z + export HIP_VISIBLE_DEVICES=0 2024-08-06T20:39:49.0843213Z + HIP_VISIBLE_DEVICES=0 2024-08-06T20:39:49.0843565Z + [[ default == \d\i\s\t\r\i\b\u\t\e\d ]] 2024-08-06T20:39:49.0843969Z + [[ default == \s\l\o\w ]] 2024-08-06T20:39:49.0844401Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *slow-gradcheck* ]] 2024-08-06T20:39:49.0844948Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *cuda* ]] 2024-08-06T20:39:49.0845434Z + export PYTORCH_TESTING_DEVICE_ONLY_FOR=cuda 2024-08-06T20:39:49.0845867Z + PYTORCH_TESTING_DEVICE_ONLY_FOR=cuda 2024-08-06T20:39:49.0846273Z + [[ default == *crossref* ]] 2024-08-06T20:39:49.0846672Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *rocm* ]] 2024-08-06T20:39:49.0847165Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *xpu* ]] 2024-08-06T20:39:49.0847673Z + [[ linux-focal-cuda12.1-py3.10-gcc9 != *-bazel-* ]] 2024-08-06T20:39:49.0848131Z + pip_install --user ninja==1.10.2 2024-08-06T20:39:49.0848587Z + pip install --progress-bar off --user ninja==1.10.2 2024-08-06T20:39:49.6958839Z Collecting ninja==1.10.2 2024-08-06T20:39:49.7160602Z Downloading ninja-1.10.2-py2.py3-none-manylinux_2_5_x86_64.manylinux1_x86_64.whl.metadata (5.0 kB) 2024-08-06T20:39:49.7360481Z Downloading ninja-1.10.2-py2.py3-none-manylinux_2_5_x86_64.manylinux1_x86_64.whl (108 kB) 2024-08-06T20:39:50.9390287Z Installing collected packages: ninja 2024-08-06T20:39:50.9494852Z  WARNING: The script ninja is installed in '/var/lib/jenkins/.local/bin' which is not on PATH. 2024-08-06T20:39:50.9496052Z Consider adding this directory to PATH or, if you prefer to suppress this warning, use --no-warn-script-location. 2024-08-06T20:39:50.9566584Z Successfully installed ninja-1.10.2 2024-08-06T20:39:51.0240921Z + export PATH=/var/lib/jenkins/.local/bin:/opt/cache/bin:/opt/conda/envs/py_3.10/bin:/opt/conda/bin:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin 2024-08-06T20:39:51.0242875Z + PATH=/var/lib/jenkins/.local/bin:/opt/cache/bin:/opt/conda/envs/py_3.10/bin:/opt/conda/bin:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin 2024-08-06T20:39:51.0244050Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *aarch64* ]] 2024-08-06T20:39:51.0244514Z + install_tlparse 2024-08-06T20:39:51.0244843Z + pip_install --user tlparse==0.3.7 2024-08-06T20:39:51.0245338Z + pip install --progress-bar off --user tlparse==0.3.7 2024-08-06T20:39:51.5546014Z Collecting tlparse==0.3.7 2024-08-06T20:39:51.5727040Z Downloading tlparse-0.3.7-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (346 bytes) 2024-08-06T20:39:51.5857736Z Downloading tlparse-0.3.7-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (2.2 MB) 2024-08-06T20:39:52.8178081Z Installing collected packages: tlparse 2024-08-06T20:39:52.8679871Z Successfully installed tlparse-0.3.7 2024-08-06T20:39:52.9354854Z ++ python -m site --user-base 2024-08-06T20:39:52.9601082Z + PATH=/var/lib/jenkins/.local/bin:/var/lib/jenkins/.local/bin:/opt/cache/bin:/opt/conda/envs/py_3.10/bin:/opt/conda/bin:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin 2024-08-06T20:39:52.9602633Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *asan* ]] 2024-08-06T20:39:52.9603152Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *-debug* ]] 2024-08-06T20:39:52.9603655Z + [[ linux-focal-cuda12.1-py3.10-gcc9 != *-bazel-* ]] 2024-08-06T20:39:52.9604622Z + echo 'We are not in debug mode: linux-focal-cuda12.1-py3.10-gcc9. Expect the assertion to pass' 2024-08-06T20:39:52.9605650Z We are not in debug mode: linux-focal-cuda12.1-py3.10-gcc9. Expect the assertion to pass 2024-08-06T20:39:52.9607029Z + cd test 2024-08-06T20:39:52.9607480Z + python -c 'import torch; torch._C._crash_if_debug_asserts_fail(424242)' 2024-08-06T20:39:55.0060296Z + [[ default == \n\o\g\p\u\_\N\O\_\A\V\X\2 ]] 2024-08-06T20:39:55.0060843Z + [[ default == \n\o\g\p\u\_\A\V\X\5\1\2 ]] 2024-08-06T20:39:55.0065773Z + DYNAMO_BENCHMARK_FLAGS=() 2024-08-06T20:39:55.0066175Z + [[ default == *dynamo_eager* ]] 2024-08-06T20:39:55.0066813Z + [[ default == *aot_eager* ]] 2024-08-06T20:39:55.0067189Z + [[ default == *aot_inductor* ]] 2024-08-06T20:39:55.0067563Z + [[ default == *inductor* ]] 2024-08-06T20:39:55.0068145Z + [[ default == *dynamic* ]] 2024-08-06T20:39:55.0068507Z + [[ default == *cpu* ]] 2024-08-06T20:39:55.0069178Z + DYNAMO_BENCHMARK_FLAGS+=(--device cuda) 2024-08-06T20:39:55.0107965Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *libtorch* ]] 2024-08-06T20:39:55.0108733Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *-bazel-* ]] 2024-08-06T20:39:55.0111573Z + cd test 2024-08-06T20:39:55.0111977Z + python -c 'import torch; print(torch.__config__.show())' 2024-08-06T20:39:56.8204554Z PyTorch built with: 2024-08-06T20:39:56.8205208Z - GCC 9.4 2024-08-06T20:39:56.8205496Z - C++ Version: 201703 2024-08-06T20:39:56.8206251Z - Intel(R) oneAPI Math Kernel Library Version 2021.4-Product Build 20210904 for Intel(R) 64 architecture applications 2024-08-06T20:39:56.8207160Z - Intel(R) MKL-DNN v3.4.2 (Git Hash 1137e04ec0b5251ca2b4400a4fd3c667ce843d67) 2024-08-06T20:39:56.8207701Z - OpenMP 201511 (a.k.a. OpenMP 4.5) 2024-08-06T20:39:56.8208139Z - LAPACK is enabled (usually provided by MKL) 2024-08-06T20:39:56.8208571Z - NNPACK is enabled 2024-08-06T20:39:56.8208900Z - CPU capability usage: AVX2 2024-08-06T20:39:56.8209265Z - CUDA Runtime 12.1 2024-08-06T20:39:56.8209716Z - NVCC architecture flags: -gencode;arch=compute_52,code=sm_52 2024-08-06T20:39:56.8210223Z - CuDNN 90.1 (built against CUDA 12.4) 2024-08-06T20:39:56.8210618Z - Magma 2.6.1 2024-08-06T20:39:56.8216405Z - Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=12.1, CUDNN_VERSION=9.1.0, CXX_COMPILER=/opt/cache/bin/c++, CXX_FLAGS= -D_GLIBCXX_USE_CXX11_ABI=1 -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -DNDEBUG -DUSE_KINETO -DLIBKINETO_NOROCTRACER -DUSE_FBGEMM -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -O2 -fPIC -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Werror=bool-operation -Wnarrowing -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-unused-parameter -Wno-strict-overflow -Wno-strict-aliasing -Wno-stringop-overflow -Wsuggest-override -Wno-psabi -Wno-error=pedantic -Wno-error=old-style-cast -Wno-missing-braces -fdiagnostics-color=always -faligned-new -Werror -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, FORCE_FALLBACK_CUDA_MPI=1, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_VERSION=2.5.0, USE_CUDA=ON, USE_CUDNN=ON, USE_CUSPARSELT=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_GLOO=ON, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=ON, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, USE_ROCM=OFF, USE_ROCM_KERNEL_ASSERT=OFF, 2024-08-06T20:39:56.8222734Z 2024-08-06T20:39:57.0732498Z + cd test 2024-08-06T20:39:57.0733008Z + python -c 'import torch; print(torch.__config__.parallel_info())' 2024-08-06T20:39:58.8562710Z ATen/Parallel: 2024-08-06T20:39:58.8563178Z at::get_num_threads() : 8 2024-08-06T20:39:58.8563563Z at::get_num_interop_threads() : 8 2024-08-06T20:39:58.8563965Z OpenMP 201511 (a.k.a. OpenMP 4.5) 2024-08-06T20:39:58.8564355Z omp_get_max_threads() : 8 2024-08-06T20:39:58.8565036Z Intel(R) oneAPI Math Kernel Library Version 2021.4-Product Build 20210904 for Intel(R) 64 architecture applications 2024-08-06T20:39:58.8565771Z mkl_get_max_threads() : 8 2024-08-06T20:39:58.8566263Z Intel(R) MKL-DNN v3.4.2 (Git Hash 1137e04ec0b5251ca2b4400a4fd3c667ce843d67) 2024-08-06T20:39:58.8566814Z std::thread::hardware_concurrency() : 16 2024-08-06T20:39:58.8567231Z Environment variables: 2024-08-06T20:39:58.8567580Z OMP_NUM_THREADS : [not set] 2024-08-06T20:39:58.8567927Z MKL_NUM_THREADS : [not set] 2024-08-06T20:39:58.8568310Z ATen parallel backend: OpenMP 2024-08-06T20:39:58.8568560Z 2024-08-06T20:39:59.0927445Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *aarch64* ]] 2024-08-06T20:39:59.0928000Z + [[ default == *backward* ]] 2024-08-06T20:39:59.0928386Z + [[ default == *xla* ]] 2024-08-06T20:39:59.0928745Z + [[ default == *executorch* ]] 2024-08-06T20:39:59.0929136Z + [[ default == \j\i\t\_\l\e\g\a\c\y ]] 2024-08-06T20:39:59.0929583Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *libtorch* ]] 2024-08-06T20:39:59.0930055Z + [[ default == distributed ]] 2024-08-06T20:39:59.0930440Z + [[ default == *inductor_distributed* ]] 2024-08-06T20:39:59.0930850Z + [[ default == *inductor-halide* ]] 2024-08-06T20:39:59.0931281Z + [[ default == *inductor-micro-benchmark* ]] 2024-08-06T20:39:59.0931721Z + [[ default == *huggingface* ]] 2024-08-06T20:39:59.0932079Z + [[ default == *timm* ]] 2024-08-06T20:39:59.0932559Z + [[ default == *torchbench* ]] 2024-08-06T20:39:59.0933246Z + [[ default == *inductor_cpp_wrapper_abi_compatible* ]] 2024-08-06T20:39:59.0933711Z + [[ default == *inductor* ]] 2024-08-06T20:39:59.0934075Z + [[ default == *dynamo* ]] 2024-08-06T20:39:59.0934487Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *rocm* ]] 2024-08-06T20:39:59.0934915Z + [[ 1 == 1 ]] 2024-08-06T20:39:59.0935225Z + [[ 5 -gt 1 ]] 2024-08-06T20:39:59.0935534Z + test_without_numpy 2024-08-06T20:39:59.0935862Z ++ dirname .ci/pytorch/test.sh 2024-08-06T20:39:59.0947299Z + pushd .ci/pytorch 2024-08-06T20:39:59.0947653Z ~/workspace/.ci/pytorch ~/workspace 2024-08-06T20:39:59.0948642Z + python -c 'import sys;sys.path.insert(0, '\''fake_numpy'\'');from unittest import TestCase;import torch;x=torch.randn(3,3);TestCase().assertRaises(RuntimeError, lambda: x.numpy())' 2024-08-06T20:40:00.2149636Z /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/_subclasses/functional_tensor.py:271: UserWarning: Failed to initialize NumPy: Sorry PyTorch, but our NumPy is in the other folder (Triggered internally at /var/lib/jenkins/workspace/torch/csrc/utils/tensor_numpy.cpp:84.) 2024-08-06T20:40:00.2151400Z cpu = _conversion_method_template(device=torch.device("cpu")) 2024-08-06T20:40:00.9687179Z + python -c 'import sys;sys.path.insert(0, '\''fake_numpy'\'');import torch;print(torch.tensor([torch.tensor(0.), torch.tensor(1.)]))' 2024-08-06T20:40:02.1030449Z /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/_subclasses/functional_tensor.py:271: UserWarning: Failed to initialize NumPy: Sorry PyTorch, but our NumPy is in the other folder (Triggered internally at /var/lib/jenkins/workspace/torch/csrc/utils/tensor_numpy.cpp:84.) 2024-08-06T20:40:02.1032161Z cpu = _conversion_method_template(device=torch.device("cpu")) 2024-08-06T20:40:02.6476891Z tensor([0., 1.]) 2024-08-06T20:40:02.8755177Z + [[ default == *dynamo* ]] 2024-08-06T20:40:02.8755586Z + popd 2024-08-06T20:40:02.8755853Z ~/workspace 2024-08-06T20:40:02.8756144Z + install_torchvision 2024-08-06T20:40:02.8756479Z + local orig_preload 2024-08-06T20:40:02.8756781Z + local commit 2024-08-06T20:40:02.8760387Z ++ get_pinned_commit vision 2024-08-06T20:40:02.8761273Z ++ cat .github/ci_commit_pins/vision.txt 2024-08-06T20:40:02.8777543Z + commit=d23a6e1664d20707c11781299611436e1f0c104f 2024-08-06T20:40:02.8778000Z + orig_preload= 2024-08-06T20:40:02.8778552Z + '[' -n '' ']' 2024-08-06T20:40:02.8779264Z + pip_install --no-use-pep517 --user git+https://github.com/pytorch/vision.git@d23a6e1664d20707c11781299611436e1f0c104f 2024-08-06T20:40:02.8780518Z + pip install --progress-bar off --no-use-pep517 --user git+https://github.com/pytorch/vision.git@d23a6e1664d20707c11781299611436e1f0c104f 2024-08-06T20:40:03.3599068Z Collecting git+https://github.com/pytorch/vision.git@d23a6e1664d20707c11781299611436e1f0c104f 2024-08-06T20:40:03.3606473Z Cloning https://github.com/pytorch/vision.git (to revision d23a6e1664d20707c11781299611436e1f0c104f) to /tmp/pip-req-build-85b03_am 2024-08-06T20:40:03.3631561Z Running command git clone --filter=blob:none --quiet https://github.com/pytorch/vision.git /tmp/pip-req-build-85b03_am 2024-08-06T20:40:05.1054669Z Running command git rev-parse -q --verify 'sha^d23a6e1664d20707c11781299611436e1f0c104f' 2024-08-06T20:40:05.1080793Z Running command git fetch -q https://github.com/pytorch/vision.git d23a6e1664d20707c11781299611436e1f0c104f 2024-08-06T20:40:06.7690905Z Running command git checkout -q d23a6e1664d20707c11781299611436e1f0c104f 2024-08-06T20:40:07.0908837Z Resolved https://github.com/pytorch/vision.git to commit d23a6e1664d20707c11781299611436e1f0c104f 2024-08-06T20:40:09.9234707Z Preparing metadata (setup.py) ... [?25l- \ done 2024-08-06T20:40:09.9298497Z [?25hRequirement already satisfied: numpy in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torchvision==0.19.0a0+d23a6e1) (1.21.2) 2024-08-06T20:40:09.9302834Z Requirement already satisfied: torch in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torchvision==0.19.0a0+d23a6e1) (2.5.0a0+gitb9d86fa) 2024-08-06T20:40:09.9312009Z Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torchvision==0.19.0a0+d23a6e1) (10.3.0) 2024-08-06T20:40:09.9644220Z Requirement already satisfied: filelock in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torch->torchvision==0.19.0a0+d23a6e1) (3.13.1) 2024-08-06T20:40:09.9652217Z Requirement already satisfied: typing-extensions>=4.8.0 in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torch->torchvision==0.19.0a0+d23a6e1) (4.12.2) 2024-08-06T20:40:09.9656946Z Requirement already satisfied: networkx in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torch->torchvision==0.19.0a0+d23a6e1) (2.8.8) 2024-08-06T20:40:09.9661648Z Requirement already satisfied: jinja2 in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torch->torchvision==0.19.0a0+d23a6e1) (3.1.4) 2024-08-06T20:40:09.9666115Z Requirement already satisfied: fsspec in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torch->torchvision==0.19.0a0+d23a6e1) (2024.6.1) 2024-08-06T20:40:09.9681751Z Requirement already satisfied: sympy>=1.13.0 in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from torch->torchvision==0.19.0a0+d23a6e1) (1.13.1) 2024-08-06T20:40:09.9728684Z Requirement already satisfied: mpmath<1.4,>=1.1.0 in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from sympy>=1.13.0->torch->torchvision==0.19.0a0+d23a6e1) (1.3.0) 2024-08-06T20:40:10.1047684Z Requirement already satisfied: MarkupSafe>=2.0 in /opt/conda/envs/py_3.10/lib/python3.10/site-packages (from jinja2->torch->torchvision==0.19.0a0+d23a6e1) (2.1.5) 2024-08-06T20:40:10.1369323Z Building wheels for collected packages: torchvision 2024-08-06T20:41:52.8740096Z Building wheel for torchvision (setup.py) ... [?25l- \ | / - \ | / - \ | / - \ | / - \ | / - \ | / - \ | / - \ | / - \ | / - \ | / - \ | done 2024-08-06T20:41:52.8825849Z [?25h Created wheel for torchvision: filename=torchvision-0.19.0a0+d23a6e1-cp310-cp310-linux_x86_64.whl size=2116262 sha256=86b514143ba4b96812c36f9aad00af26736f0b0b8f6f0d14425603b423c3ca12 2024-08-06T20:41:52.8829331Z Stored in directory: /var/lib/jenkins/.cache/pip/wheels/0e/56/35/02931e71eb23fd2b85591c7ec05b733ca7c8b328a2fd151f96 2024-08-06T20:41:52.8877525Z Successfully built torchvision 2024-08-06T20:41:53.8799381Z Installing collected packages: torchvision 2024-08-06T20:41:54.4467247Z Successfully installed torchvision-0.19.0a0+d23a6e1 2024-08-06T20:41:54.5635730Z + '[' -n '' ']' 2024-08-06T20:41:54.5636128Z + test_python_shard 1 2024-08-06T20:41:54.5636467Z + [[ -z 5 ]] 2024-08-06T20:41:54.5637351Z + python test/run_test.py --exclude-jit-executor --exclude-distributed-tests --shard 1 5 --verbose 2024-08-06T20:41:54.6960123Z /var/lib/jenkins/workspace/test/run_test.py:21: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html 2024-08-06T20:41:54.6961182Z import pkg_resources 2024-08-06T20:41:59.0816981Z Downloading https://ossci-metrics.s3.amazonaws.com/slow-tests.json to /var/lib/jenkins/workspace/test/.pytorch-slow-tests.json 2024-08-06T20:41:59.1433544Z Downloading https://ossci-metrics.s3.amazonaws.com/disabled-tests-condensed.json to /var/lib/jenkins/workspace/test/.pytorch-disabled-tests.json 2024-08-06T20:41:59.1857035Z Ignoring disabled issues: [''] 2024-08-06T20:41:59.1994871Z Found test times from artifacts 2024-08-06T20:41:59.2523097Z Found test times from artifacts 2024-08-06T20:41:59.2538537Z Running 25% of tests based on TD 2024-08-06T20:41:59.2875654Z Running parallel tests on 2 processes 2024-08-06T20:41:59.2880237Z Name: tests to run (est. time: 73.79min) 2024-08-06T20:41:59.2880700Z Serial tests (6): 2024-08-06T20:41:59.2881022Z test_native_mha 1/1 2024-08-06T20:41:59.2881571Z test_nn 1/1 2024-08-06T20:41:59.2881868Z test_overrides 1/1 2024-08-06T20:41:59.2882217Z test_cpp_api_parity 1/1 2024-08-06T20:41:59.2882797Z test_torch 1/1 2024-08-06T20:41:59.2883122Z test_namedtuple_return_api 1/1 2024-08-06T20:41:59.2883516Z Parallel tests (41): 2024-08-06T20:41:59.2884091Z test_decomp 5/24 2024-08-06T20:41:59.2884411Z test_decomp 14/24 2024-08-06T20:41:59.2884740Z test_decomp 15/24 2024-08-06T20:41:59.2885185Z test_decomp 24/24 2024-08-06T20:41:59.2885634Z test_model_exports_to_core_aten 1/1 2024-08-06T20:41:59.2886044Z test_transformers 1/1 2024-08-06T20:41:59.2886534Z functorch/test_ops 8/9 2024-08-06T20:41:59.2886989Z test_proxy_tensor 1/1 2024-08-06T20:41:59.2887330Z test_foreach 1/1 2024-08-06T20:41:59.2887730Z test_ops_gradients 1/4 2024-08-06T20:41:59.2888229Z test_meta 5/5 2024-08-06T20:41:59.2888534Z test_ops 2/11 2024-08-06T20:41:59.2888820Z test_ops 9/11 2024-08-06T20:41:59.2889349Z test_module_tracker 1/1 2024-08-06T20:41:59.2889742Z test_comparison_utils 1/1 2024-08-06T20:41:59.2890365Z functorch/test_vmap 1/1 2024-08-06T20:41:59.2890906Z inductor/test_inplacing_pass 1/1 2024-08-06T20:41:59.2891310Z export/test_tools 1/1 2024-08-06T20:41:59.2891863Z inductor/test_torchinductor_dynamic_shapes 1/4 2024-08-06T20:41:59.2892417Z torch_np/test_binary_ufuncs 1/1 2024-08-06T20:41:59.2892848Z dynamo/test_inline_inbuilt_nn_modules 1/1 2024-08-06T20:41:59.2893514Z inductor/test_inductor_utils 1/1 2024-08-06T20:41:59.2893918Z torch_np/test_unary_ufuncs 1/1 2024-08-06T20:41:59.2894479Z inductor/test_extension_backend 1/1 2024-08-06T20:41:59.2894965Z export/test_export_nonstrict 1/1 2024-08-06T20:41:59.2895412Z inductor/test_triton_extension_backend 1/1 2024-08-06T20:41:59.2896093Z inductor/test_triton_kernels 1/1 2024-08-06T20:41:59.2896493Z dynamo/test_base_output 1/1 2024-08-06T20:41:59.2896961Z export/test_serdes 1/1 2024-08-06T20:41:59.2897460Z dynamo/test_interop 1/1 2024-08-06T20:41:59.2897813Z inductor/test_compile_worker 1/1 2024-08-06T20:41:59.2898363Z dynamo/test_fx_passes_pre_grad 1/1 2024-08-06T20:41:59.2898866Z inductor/test_flex_decoding 1/1 2024-08-06T20:41:59.2899250Z dynamo/test_skip_non_tensor 1/1 2024-08-06T20:41:59.2900248Z inductor/test_move_constructors_to_cuda 1/1 2024-08-06T20:41:59.2900687Z test_matmul_cuda 1/1 2024-08-06T20:41:59.2901258Z dynamo/test_sdpa 1/1 2024-08-06T20:41:59.2901822Z inductor/test_metrics 1/1 2024-08-06T20:41:59.2902393Z inductor/test_debug_trace 1/1 2024-08-06T20:41:59.2902849Z test_modules 1/2 2024-08-06T20:41:59.2903199Z inductor/test_cuda_cpp_wrapper 1/1 2024-08-06T20:41:59.2903845Z Name: excluded (est. time: 32.15min) 2024-08-06T20:41:59.2904504Z Serial tests (39): 2024-08-06T20:41:59.2905052Z test_fx 1/1 2024-08-06T20:41:59.2905491Z test_multiprocessing_spawn 1/1 2024-08-06T20:41:59.2905878Z test_cpp_extensions_jit 1/1 2024-08-06T20:41:59.2906385Z test_tensorexpr 1/1 2024-08-06T20:41:59.2906820Z test_reductions 1/1 2024-08-06T20:41:59.2907151Z test_multiprocessing 1/1 2024-08-06T20:41:59.2907618Z test_tensor_creation_ops 1/1 2024-08-06T20:41:59.2908158Z distributions/test_distributions 1/1 2024-08-06T20:41:59.2908596Z inductor/test_distributed_patterns 1/1 2024-08-06T20:41:59.2909246Z nn/test_convolution 1/1 2024-08-06T20:41:59.2909609Z test_dispatch 1/1 2024-08-06T20:41:59.2909921Z test_utils 1/1 2024-08-06T20:41:59.2910430Z test_fake_tensor 1/1 2024-08-06T20:41:59.2910809Z test_spectral_ops 1/1 2024-08-06T20:41:59.2911155Z inductor/test_benchmark_fusion 1/1 2024-08-06T20:41:59.2911745Z nn/test_pooling 1/1 2024-08-06T20:41:59.2912174Z test_cpp_extensions_open_device_registration 1/1 2024-08-06T20:41:59.2912659Z functorch/test_memory_efficient_fusion 1/1 2024-08-06T20:41:59.2913317Z test_sort_and_select 1/1 2024-08-06T20:41:59.2913688Z test_mobile_optimizer 1/1 2024-08-06T20:41:59.2914133Z test_python_dispatch 1/1 2024-08-06T20:41:59.2914638Z test_autocast 1/1 2024-08-06T20:41:59.2914946Z test_cuda_trace 1/1 2024-08-06T20:41:59.2915319Z test_autograd_fallback 1/1 2024-08-06T20:41:59.2915883Z test_jit_disabled 1/1 2024-08-06T20:41:59.2916244Z test_cpp_extensions_aot_ninja 1/1 2024-08-06T20:41:59.2916773Z test_cpp_extensions_aot_no_ninja 1/1 2024-08-06T20:41:59.2917314Z test_show_pickle 1/1 2024-08-06T20:41:59.2917661Z test_cuda_nvml_based_avail 1/1 2024-08-06T20:41:59.2918240Z inductor/test_flex_attention 1/1 2024-08-06T20:41:59.2918698Z test_cpp_extensions_mtia_backend 1/1 2024-08-06T20:41:59.2919094Z test_cuda_primary_ctx 1/1 2024-08-06T20:41:59.2919718Z test_cpp_extensions_stream_and_event 1/1 2024-08-06T20:41:59.2920159Z inductor/test_cutlass_backend 1/1 2024-08-06T20:41:59.2920628Z test_ci_sanity_check_fail 1/1 2024-08-06T20:41:59.2921173Z inductor/test_max_autotune 1/1 2024-08-06T20:41:59.2921554Z doctests 1/1 2024-08-06T20:41:59.2921952Z test_autoload_enable 1/1 2024-08-06T20:41:59.2922457Z test_autoload_disable 1/1 2024-08-06T20:41:59.2922824Z Parallel tests (55): 2024-08-06T20:41:59.2923230Z test_schema_check 1/1 2024-08-06T20:41:59.2923732Z inductor/test_minifier 1/1 2024-08-06T20:41:59.2924105Z test_masked 1/1 2024-08-06T20:41:59.2924453Z export/test_converter 1/1 2024-08-06T20:41:59.2925019Z dynamo/test_hooks 1/1 2024-08-06T20:41:59.2925422Z torch_np/numpy_tests/core/test_scalarmath 1/1 2024-08-06T20:41:59.2926034Z functorch/test_dims 1/1 2024-08-06T20:41:59.2926445Z test_tensorboard 1/1 2024-08-06T20:41:59.2926800Z dynamo/test_aot_autograd 1/1 2024-08-06T20:41:59.2927352Z test_xnnpack_integration 1/1 2024-08-06T20:41:59.2927766Z test_jiterator 1/1 2024-08-06T20:41:59.2928087Z dynamo/test_modules 1/1 2024-08-06T20:41:59.2928602Z dynamo/test_logging 1/1 2024-08-06T20:41:59.2929032Z test_sympy_utils 1/1 2024-08-06T20:41:59.2929367Z export/test_passes 1/1 2024-08-06T20:41:59.2929838Z test_datapipe 1/1 2024-08-06T20:41:59.2930298Z dynamo/test_input_attr_tracking 1/1 2024-08-06T20:41:59.2930697Z dynamo/test_cudagraphs 1/1 2024-08-06T20:41:59.2931303Z torch_np/numpy_tests/fft/test_pocketfft 1/1 2024-08-06T20:41:59.2932015Z functorch/test_vmap_registrations 1/1 2024-08-06T20:41:59.2932611Z test_logging 1/1 2024-08-06T20:41:59.2932990Z dynamo/test_minifier 1/1 2024-08-06T20:41:59.2933478Z test_indexing 1/1 2024-08-06T20:41:59.2934051Z inductor/test_ordered_set 1/1 2024-08-06T20:41:59.2934451Z dynamo/test_recompile_ux 1/1 2024-08-06T20:41:59.2934826Z test_model_dump 1/1 2024-08-06T20:41:59.2935376Z dynamo/test_sources 1/1 2024-08-06T20:41:59.2935740Z test_out_dtype_op 1/1 2024-08-06T20:41:59.2936084Z dynamo/test_view 1/1 2024-08-06T20:41:59.2936740Z dynamo/test_activation_checkpointing 1/1 2024-08-06T20:41:59.2937327Z nn/test_packed_sequence 1/1 2024-08-06T20:41:59.2938162Z test_flop_counter 1/1 2024-08-06T20:41:59.2938642Z nn/test_lazy_modules 1/1 2024-08-06T20:41:59.2939329Z test_monitor 1/1 2024-08-06T20:41:59.2939853Z dynamo/test_modes 1/1 2024-08-06T20:41:59.2940534Z profiler/test_profiler_tree 1/1 2024-08-06T20:41:59.2941094Z functorch/test_logging 1/1 2024-08-06T20:41:59.2954402Z export/test_schema 1/1 2024-08-06T20:41:59.2954836Z test_set_default_mobile_cpu_allocator 1/1 2024-08-06T20:41:59.2955285Z torch_np/test_function_base 1/1 2024-08-06T20:41:59.2955821Z nn/test_pruning 1/1 2024-08-06T20:41:59.2956325Z test_jit_llga_fuser 1/1 2024-08-06T20:41:59.2956684Z inductor/test_control_flow 1/1 2024-08-06T20:41:59.2957080Z inductor/test_mmdecomp 1/1 2024-08-06T20:41:59.2957470Z dynamo/test_deviceguard 1/1 2024-08-06T20:41:59.2957865Z inductor/test_graph_transform_observer 1/1 2024-08-06T20:41:59.2958283Z dynamo/test_model_output 1/1 2024-08-06T20:41:59.2958730Z torch_np/numpy_tests/core/test_scalar_methods 1/1 2024-08-06T20:41:59.2959182Z inductor/test_triton_wrapper 1/1 2024-08-06T20:41:59.2959591Z inductor/test_pattern_matcher 1/1 2024-08-06T20:41:59.2959999Z inductor/test_scatter_optimization 1/1 2024-08-06T20:41:59.2960428Z inductor/test_halide 1/1 2024-08-06T20:41:59.2960831Z inductor/test_loop_ordering 1/1 2024-08-06T20:41:59.2961216Z inductor/test_unbacked_symints 1/1 2024-08-06T20:41:59.2961636Z optim/test_swa_utils 1/1 2024-08-06T20:41:59.2962146Z Starting test batch 'tests to run' 0.0 seconds after initiating testing 2024-08-06T20:41:59.2962792Z Running test_native_mha 1/1 ... [2024-08-06 20:41:59.295690] 2024-08-06T20:41:59.2965916Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_native_mha.py', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:41:59.296185] 2024-08-06T20:42:04.4212841Z 2024-08-06T20:42:04.4213731Z test_native_mha 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_native_mha_1.1_43f467f7830589b4_.log 2024-08-06T20:42:04.4263389Z Running 54 items in this shard: test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_attention_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_attention_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_encoder_decoder_attention_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_encoder_decoder_attention_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_False_fused_False_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_False_fused_False_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_False_fused_True_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_False_fused_True_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_True_fused_False_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_True_fused_False_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_True_fused_True_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_True_fused_True_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_False_fused_False_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_False_fused_False_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_False_fused_True_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_False_fused_True_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_True_fused_False_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_True_fused_False_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_True_fused_True_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_True_fused_True_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_False_fused_False_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_False_fused_False_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_False_fused_True_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_False_fused_True_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_True_fused_False_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_True_fused_False_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_True_fused_True_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_False_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_True_fused_True_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_False_fused_False_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_False_fused_False_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_False_fused_True_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_False_fused_True_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_True_fused_False_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_True_fused_False_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_True_fused_True_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_False_pad_all_False_need_weights_False_average_attn_weights_True_fused_True_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_False_fused_False_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_False_fused_False_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_False_fused_True_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_False_fused_True_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_True_fused_False_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_True_fused_False_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_True_fused_True_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_False_need_weights_False_average_attn_weights_True_fused_True_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_False_fused_False_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_False_fused_False_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_False_fused_True_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_False_fused_True_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_True_fused_False_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_True_fused_False_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_True_fused_True_cuda_float16, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_native_multihead_self_attention_use_nt_True_use_padding_True_pad_all_True_need_weights_False_average_attn_weights_True_fused_True_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_transform_bias_rescale_qkv_cuda_float32, test/test_native_mha.py::TestMHADeviceTypeCUDA::test_transform_bias_rescale_qkv_nested_cuda_float32 2024-08-06T20:42:04.4309132Z 2024-08-06T20:42:04.4309331Z Running test_nn 1/1 ... [2024-08-06 20:42:04.421935] 2024-08-06T20:42:04.4310555Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_nn.py', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:42:04.422398] 2024-08-06T20:45:29.4824973Z 2024-08-06T20:45:29.4825872Z test_nn 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_nn_1.1_5430fad9bab7e294_.log 2024-08-06T20:45:29.6177402Z Running 2195 items in this shard: test/test_nn.py::TestNN::test_AdaptiveLogSoftmax, test/test_nn.py::TestNN::test_AdaptiveLogSoftmax_cuda, test/test_nn.py::TestNN::test_BCELoss_no_batch_dim_mean, test/test_nn.py::TestNN::test_BCELoss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_BCELoss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_BCELoss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_BCELoss_no_batch_dim_none, test/test_nn.py::TestNN::test_BCELoss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_BCELoss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_BCELoss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_BCELoss_no_batch_dim_sum, test/test_nn.py::TestNN::test_BCELoss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_BCELoss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_BCELoss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_BCELoss_no_reduce, test/test_nn.py::TestNN::test_BCELoss_no_reduce_cuda, test/test_nn.py::TestNN::test_BCELoss_no_reduce_scalar, test/test_nn.py::TestNN::test_BCELoss_no_reduce_scalar_cuda, test/test_nn.py::TestNN::test_BCELoss_weights_no_reduce, test/test_nn.py::TestNN::test_BCELoss_weights_no_reduce_cuda, test/test_nn.py::TestNN::test_BCELoss_weights_no_reduce_scalar, test/test_nn.py::TestNN::test_BCELoss_weights_no_reduce_scalar_cuda, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_legacy_enum, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_legacy_enum_cuda, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_batch_dim_mean, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_batch_dim_none, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_batch_dim_sum, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_reduce, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_reduce_cuda, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_reduce_scalar, test/test_nn.py::TestNN::test_BCEWithLogitsLoss_no_reduce_scalar_cuda, test/test_nn.py::TestNN::test_CELU_no_batch_dim, test/test_nn.py::TestNN::test_CELU_no_batch_dim_cuda, test/test_nn.py::TestNN::test_CTCLoss_critical_target_len, test/test_nn.py::TestNN::test_CTCLoss_lengthchecks_cpu, test/test_nn.py::TestNN::test_CTCLoss_lengthchecks_cuda, test/test_nn.py::TestNN::test_CTCLoss_long_targets, test/test_nn.py::TestNN::test_CTCLoss_typechecks, test/test_nn.py::TestNN::test_CTCLoss_zero_infinity, test/test_nn.py::TestNN::test_CTCLoss_zero_lengths, test/test_nn.py::TestNN::test_Conv1d, test/test_nn.py::TestNN::test_Conv1d_circular_stride2_pad2, test/test_nn.py::TestNN::test_Conv1d_circular_stride2_pad2_cuda, test/test_nn.py::TestNN::test_Conv1d_cuda, test/test_nn.py::TestNN::test_Conv1d_dilated, test/test_nn.py::TestNN::test_Conv1d_dilated_cuda, test/test_nn.py::TestNN::test_Conv1d_groups, test/test_nn.py::TestNN::test_Conv1d_groups_cuda, test/test_nn.py::TestNN::test_Conv1d_pad1, test/test_nn.py::TestNN::test_Conv1d_pad1_cuda, test/test_nn.py::TestNN::test_Conv1d_pad1size1, test/test_nn.py::TestNN::test_Conv1d_pad1size1_cuda, test/test_nn.py::TestNN::test_Conv1d_pad2, test/test_nn.py::TestNN::test_Conv1d_pad2_cuda, test/test_nn.py::TestNN::test_Conv1d_pad2size1, test/test_nn.py::TestNN::test_Conv1d_pad2size1_cuda, test/test_nn.py::TestNN::test_Conv1d_pad_same, test/test_nn.py::TestNN::test_Conv1d_pad_same2, test/test_nn.py::TestNN::test_Conv1d_pad_same2_cuda, test/test_nn.py::TestNN::test_Conv1d_pad_same_cuda, test/test_nn.py::TestNN::test_Conv1d_pad_same_dilated, test/test_nn.py::TestNN::test_Conv1d_pad_same_dilated_cuda, test/test_nn.py::TestNN::test_Conv1d_pad_valid, test/test_nn.py::TestNN::test_Conv1d_pad_valid_cuda, test/test_nn.py::TestNN::test_Conv1d_reflect_stride2_pad2, test/test_nn.py::TestNN::test_Conv1d_reflect_stride2_pad2_cuda, test/test_nn.py::TestNN::test_Conv1d_replicate_stride2_pad2, test/test_nn.py::TestNN::test_Conv1d_replicate_stride2_pad2_cuda, test/test_nn.py::TestNN::test_Conv1d_stride, test/test_nn.py::TestNN::test_Conv1d_stride_cuda, test/test_nn.py::TestNN::test_Conv1d_zero_batch, test/test_nn.py::TestNN::test_Conv1d_zero_batch_cuda, test/test_nn.py::TestNN::test_Conv1d_zeros_stride2_pad2, test/test_nn.py::TestNN::test_Conv1d_zeros_stride2_pad2_cuda, test/test_nn.py::TestNN::test_Conv2d, test/test_nn.py::TestNN::test_Conv2d_circular_stride2_pad2, test/test_nn.py::TestNN::test_Conv2d_circular_stride2_pad2_cuda, test/test_nn.py::TestNN::test_Conv2d_cuda, test/test_nn.py::TestNN::test_Conv2d_depthwise, test/test_nn.py::TestNN::test_Conv2d_depthwise_cuda, test/test_nn.py::TestNN::test_Conv2d_depthwise_dilated, test/test_nn.py::TestNN::test_Conv2d_depthwise_dilated_cuda, test/test_nn.py::TestNN::test_Conv2d_depthwise_padded, test/test_nn.py::TestNN::test_Conv2d_depthwise_padded_cuda, test/test_nn.py::TestNN::test_Conv2d_depthwise_strided, test/test_nn.py::TestNN::test_Conv2d_depthwise_strided_cuda, test/test_nn.py::TestNN::test_Conv2d_depthwise_with_multiplier, test/test_nn.py::TestNN::test_Conv2d_depthwise_with_multiplier_cuda, test/test_nn.py::TestNN::test_Conv2d_dilated, test/test_nn.py::TestNN::test_Conv2d_dilated_cuda, test/test_nn.py::TestNN::test_Conv2d_dilated_with_long_tensor, test/test_nn.py::TestNN::test_Conv2d_dilated_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv2d_groups, test/test_nn.py::TestNN::test_Conv2d_groups_cuda, test/test_nn.py::TestNN::test_Conv2d_groups_thnn, test/test_nn.py::TestNN::test_Conv2d_groups_thnn_cuda, test/test_nn.py::TestNN::test_Conv2d_groups_thnn_with_long_tensor, test/test_nn.py::TestNN::test_Conv2d_groups_thnn_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv2d_groups_with_long_tensor, test/test_nn.py::TestNN::test_Conv2d_groups_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv2d_no_bias, test/test_nn.py::TestNN::test_Conv2d_no_bias_cuda, test/test_nn.py::TestNN::test_Conv2d_no_bias_with_long_tensor, test/test_nn.py::TestNN::test_Conv2d_no_bias_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv2d_pad_same, test/test_nn.py::TestNN::test_Conv2d_pad_same_cuda, test/test_nn.py::TestNN::test_Conv2d_pad_same_dilated, test/test_nn.py::TestNN::test_Conv2d_pad_same_dilated_cuda, test/test_nn.py::TestNN::test_Conv2d_pad_valid, test/test_nn.py::TestNN::test_Conv2d_pad_valid_cuda, test/test_nn.py::TestNN::test_Conv2d_padding, test/test_nn.py::TestNN::test_Conv2d_padding_cuda, test/test_nn.py::TestNN::test_Conv2d_padding_with_long_tensor, test/test_nn.py::TestNN::test_Conv2d_padding_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv2d_reflect_stride2_pad2, test/test_nn.py::TestNN::test_Conv2d_reflect_stride2_pad2_cuda, test/test_nn.py::TestNN::test_Conv2d_replicate_stride2_pad2, test/test_nn.py::TestNN::test_Conv2d_replicate_stride2_pad2_cuda, test/test_nn.py::TestNN::test_Conv2d_strided, test/test_nn.py::TestNN::test_Conv2d_strided_cuda, test/test_nn.py::TestNN::test_Conv2d_strided_with_long_tensor, test/test_nn.py::TestNN::test_Conv2d_strided_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv2d_with_long_tensor, test/test_nn.py::TestNN::test_Conv2d_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv2d_zero_batch, test/test_nn.py::TestNN::test_Conv2d_zero_batch_cuda, test/test_nn.py::TestNN::test_Conv2d_zero_batch_with_long_tensor, test/test_nn.py::TestNN::test_Conv2d_zero_batch_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv2d_zeros_stride2_pad2, test/test_nn.py::TestNN::test_Conv2d_zeros_stride2_pad2_cuda, test/test_nn.py::TestNN::test_Conv3d, test/test_nn.py::TestNN::test_Conv3d_1x1x1_no_bias, test/test_nn.py::TestNN::test_Conv3d_1x1x1_no_bias_cuda, test/test_nn.py::TestNN::test_Conv3d_1x1x1_no_bias_with_long_tensor, test/test_nn.py::TestNN::test_Conv3d_1x1x1_no_bias_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv3d_circular_stride2_pad2, test/test_nn.py::TestNN::test_Conv3d_circular_stride2_pad2_cuda, test/test_nn.py::TestNN::test_Conv3d_cuda, test/test_nn.py::TestNN::test_Conv3d_dilated, test/test_nn.py::TestNN::test_Conv3d_dilated_cuda, test/test_nn.py::TestNN::test_Conv3d_dilated_strided, test/test_nn.py::TestNN::test_Conv3d_dilated_strided_cuda, test/test_nn.py::TestNN::test_Conv3d_groups, test/test_nn.py::TestNN::test_Conv3d_groups_cuda, test/test_nn.py::TestNN::test_Conv3d_groups_with_long_tensor, test/test_nn.py::TestNN::test_Conv3d_groups_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv3d_no_bias, test/test_nn.py::TestNN::test_Conv3d_no_bias_cuda, test/test_nn.py::TestNN::test_Conv3d_no_bias_with_long_tensor, test/test_nn.py::TestNN::test_Conv3d_no_bias_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv3d_pad_same, test/test_nn.py::TestNN::test_Conv3d_pad_same_cuda, test/test_nn.py::TestNN::test_Conv3d_pad_same_dilated, test/test_nn.py::TestNN::test_Conv3d_pad_same_dilated_cuda, test/test_nn.py::TestNN::test_Conv3d_pad_valid, test/test_nn.py::TestNN::test_Conv3d_pad_valid_cuda, test/test_nn.py::TestNN::test_Conv3d_replicate_stride2_pad2, test/test_nn.py::TestNN::test_Conv3d_replicate_stride2_pad2_cuda, test/test_nn.py::TestNN::test_Conv3d_stride, test/test_nn.py::TestNN::test_Conv3d_stride_cuda, test/test_nn.py::TestNN::test_Conv3d_stride_padding, test/test_nn.py::TestNN::test_Conv3d_stride_padding_cuda, test/test_nn.py::TestNN::test_Conv3d_stride_padding_with_long_tensor, test/test_nn.py::TestNN::test_Conv3d_stride_padding_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv3d_stride_with_long_tensor, test/test_nn.py::TestNN::test_Conv3d_stride_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv3d_with_long_tensor, test/test_nn.py::TestNN::test_Conv3d_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv3d_zero_batch, test/test_nn.py::TestNN::test_Conv3d_zero_batch_cuda, test/test_nn.py::TestNN::test_Conv3d_zero_batch_with_long_tensor, test/test_nn.py::TestNN::test_Conv3d_zero_batch_with_long_tensor_cuda, test/test_nn.py::TestNN::test_Conv3d_zeros_stride2_pad2, test/test_nn.py::TestNN::test_Conv3d_zeros_stride2_pad2_cuda, test/test_nn.py::TestNN::test_ConvTranspose1d, test/test_nn.py::TestNN::test_ConvTranspose1d_cuda, test/test_nn.py::TestNN::test_ConvTranspose1d_dilated, test/test_nn.py::TestNN::test_ConvTranspose1d_dilated_cuda, test/test_nn.py::TestNN::test_ConvTranspose1d_groups, test/test_nn.py::TestNN::test_ConvTranspose1d_groups_cuda, test/test_nn.py::TestNN::test_ConvTranspose1d_no_bias, test/test_nn.py::TestNN::test_ConvTranspose1d_no_bias_cuda, test/test_nn.py::TestNN::test_ConvTranspose2d, test/test_nn.py::TestNN::test_ConvTranspose2d_cuda, test/test_nn.py::TestNN::test_ConvTranspose2d_dilated, test/test_nn.py::TestNN::test_ConvTranspose2d_dilated_cuda, test/test_nn.py::TestNN::test_ConvTranspose2d_dilated_with_long_tensor, test/test_nn.py::TestNN::test_ConvTranspose2d_dilated_with_long_tensor_cuda, test/test_nn.py::TestNN::test_ConvTranspose2d_groups, test/test_nn.py::TestNN::test_ConvTranspose2d_groups_cuda, test/test_nn.py::TestNN::test_ConvTranspose2d_groups_with_long_tensor, test/test_nn.py::TestNN::test_ConvTranspose2d_groups_with_long_tensor_cuda, test/test_nn.py::TestNN::test_ConvTranspose2d_no_bias, test/test_nn.py::TestNN::test_ConvTranspose2d_no_bias_cuda, test/test_nn.py::TestNN::test_ConvTranspose2d_no_bias_with_long_tensor, test/test_nn.py::TestNN::test_ConvTranspose2d_no_bias_with_long_tensor_cuda, test/test_nn.py::TestNN::test_ConvTranspose2d_with_long_tensor, test/test_nn.py::TestNN::test_ConvTranspose2d_with_long_tensor_cuda, test/test_nn.py::TestNN::test_ConvTranspose3d, test/test_nn.py::TestNN::test_ConvTranspose3d_cuda, test/test_nn.py::TestNN::test_ConvTranspose3d_dilated, test/test_nn.py::TestNN::test_ConvTranspose3d_dilated_cuda, test/test_nn.py::TestNN::test_CosineEmbeddingLoss_no_batch_dim_mean, test/test_nn.py::TestNN::test_CosineEmbeddingLoss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_CosineEmbeddingLoss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_CosineEmbeddingLoss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_CosineEmbeddingLoss_no_batch_dim_none, test/test_nn.py::TestNN::test_CosineEmbeddingLoss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_CosineEmbeddingLoss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_CosineEmbeddingLoss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_CosineEmbeddingLoss_no_batch_dim_sum, test/test_nn.py::TestNN::test_CosineEmbeddingLoss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_CosineEmbeddingLoss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_CosineEmbeddingLoss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_CrossMapLRN2d, test/test_nn.py::TestNN::test_CrossMapLRN2d_cuda, test/test_nn.py::TestNN::test_ELU_no_batch_dim, test/test_nn.py::TestNN::test_ELU_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Embedding, test/test_nn.py::TestNN::test_EmbeddingBag_discontiguous, test/test_nn.py::TestNN::test_EmbeddingBag_discontiguous_cuda, test/test_nn.py::TestNN::test_EmbeddingBag_max, test/test_nn.py::TestNN::test_EmbeddingBag_max_cuda, test/test_nn.py::TestNN::test_EmbeddingBag_max_padding_idx, test/test_nn.py::TestNN::test_EmbeddingBag_max_padding_idx_cuda, test/test_nn.py::TestNN::test_EmbeddingBag_mean, test/test_nn.py::TestNN::test_EmbeddingBag_mean_cuda, test/test_nn.py::TestNN::test_EmbeddingBag_mean_padding_idx, test/test_nn.py::TestNN::test_EmbeddingBag_mean_padding_idx_cuda, test/test_nn.py::TestNN::test_EmbeddingBag_sparse, test/test_nn.py::TestNN::test_EmbeddingBag_sparse_cuda, test/test_nn.py::TestNN::test_EmbeddingBag_sum, test/test_nn.py::TestNN::test_EmbeddingBag_sum_cuda, test/test_nn.py::TestNN::test_EmbeddingBag_sum_padding_idx, test/test_nn.py::TestNN::test_EmbeddingBag_sum_padding_idx_cuda, test/test_nn.py::TestNN::test_Embedding_cuda, test/test_nn.py::TestNN::test_Embedding_discontiguous, test/test_nn.py::TestNN::test_Embedding_discontiguous_cuda, test/test_nn.py::TestNN::test_Embedding_sparse, test/test_nn.py::TestNN::test_Embedding_sparse_cuda, test/test_nn.py::TestNN::test_Flatten, test/test_nn.py::TestNN::test_Flatten_cuda, test/test_nn.py::TestNN::test_Flatten_no_batch_dim, test/test_nn.py::TestNN::test_Flatten_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Fold, test/test_nn.py::TestNN::test_Fold_cuda, test/test_nn.py::TestNN::test_Fold_int_input, test/test_nn.py::TestNN::test_Fold_int_input_cuda, test/test_nn.py::TestNN::test_Fold_no_batch_dim_input, test/test_nn.py::TestNN::test_Fold_no_batch_dim_input_cuda, test/test_nn.py::TestNN::test_Fold_no_batch_dim_int_input, test/test_nn.py::TestNN::test_Fold_no_batch_dim_int_input_cuda, test/test_nn.py::TestNN::test_GELU_no_batch_dim, test/test_nn.py::TestNN::test_GELU_no_batch_dim_cuda, test/test_nn.py::TestNN::test_GLU_no_batch_dim, test/test_nn.py::TestNN::test_GLU_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Hardshrink_no_batch_dim, test/test_nn.py::TestNN::test_Hardshrink_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Hardsigmoid_no_batch_dim, test/test_nn.py::TestNN::test_Hardsigmoid_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Hardswish_no_batch_dim, test/test_nn.py::TestNN::test_Hardswish_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Hardtanh_no_batch_dim, test/test_nn.py::TestNN::test_Hardtanh_no_batch_dim_cuda, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_margin_no_reduce, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_margin_no_reduce_cuda, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_no_batch_dim_mean, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_no_batch_dim_none, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_no_batch_dim_sum, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_no_reduce, test/test_nn.py::TestNN::test_HingeEmbeddingLoss_no_reduce_cuda, test/test_nn.py::TestNN::test_HuberLoss_delta, test/test_nn.py::TestNN::test_HuberLoss_delta_cuda, test/test_nn.py::TestNN::test_HuberLoss_no_batch_dim_mean, test/test_nn.py::TestNN::test_HuberLoss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_HuberLoss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_HuberLoss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_HuberLoss_no_batch_dim_none, test/test_nn.py::TestNN::test_HuberLoss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_HuberLoss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_HuberLoss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_HuberLoss_no_batch_dim_sum, test/test_nn.py::TestNN::test_HuberLoss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_HuberLoss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_HuberLoss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_KLDivLoss_batch_mean, test/test_nn.py::TestNN::test_KLDivLoss_batch_mean_log_target, test/test_nn.py::TestNN::test_KLDivLoss_no_batch_dim_mean, test/test_nn.py::TestNN::test_KLDivLoss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_KLDivLoss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_KLDivLoss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_KLDivLoss_no_batch_dim_none, test/test_nn.py::TestNN::test_KLDivLoss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_KLDivLoss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_KLDivLoss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_KLDivLoss_no_batch_dim_sum, test/test_nn.py::TestNN::test_KLDivLoss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_KLDivLoss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_KLDivLoss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_KLDivLoss_no_reduce, test/test_nn.py::TestNN::test_KLDivLoss_no_reduce_cuda, test/test_nn.py::TestNN::test_KLDivLoss_no_reduce_log_target, test/test_nn.py::TestNN::test_KLDivLoss_no_reduce_log_target_cuda, test/test_nn.py::TestNN::test_KLDivLoss_no_reduce_scalar, test/test_nn.py::TestNN::test_KLDivLoss_no_reduce_scalar_cuda, test/test_nn.py::TestNN::test_KLDivLoss_no_reduce_scalar_log_target, test/test_nn.py::TestNN::test_KLDivLoss_no_reduce_scalar_log_target_cuda, test/test_nn.py::TestNN::test_KLDivLoss_with_log_target_no_reduce, test/test_nn.py::TestNN::test_KLDivLoss_with_log_target_no_reduce_cuda, test/test_nn.py::TestNN::test_KLDivLoss_with_target_no_reduce, test/test_nn.py::TestNN::test_KLDivLoss_with_target_no_reduce_cuda, test/test_nn.py::TestNN::test_L1Loss_no_batch_dim_mean, test/test_nn.py::TestNN::test_L1Loss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_L1Loss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_L1Loss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_L1Loss_no_batch_dim_none, test/test_nn.py::TestNN::test_L1Loss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_L1Loss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_L1Loss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_L1Loss_no_batch_dim_sum, test/test_nn.py::TestNN::test_L1Loss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_L1Loss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_L1Loss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_L1Loss_no_reduce, test/test_nn.py::TestNN::test_L1Loss_no_reduce_complex, test/test_nn.py::TestNN::test_L1Loss_no_reduce_complex_cuda, test/test_nn.py::TestNN::test_L1Loss_no_reduce_cuda, test/test_nn.py::TestNN::test_L1Loss_no_reduce_scalar, test/test_nn.py::TestNN::test_L1Loss_no_reduce_scalar_cuda, test/test_nn.py::TestNN::test_LSTM_cell, test/test_nn.py::TestNN::test_LSTM_cell_forward_hidden_size, test/test_nn.py::TestNN::test_LSTM_cell_forward_input_size, test/test_nn.py::TestNN::test_LayerNorm_3d_no_affine_large_feature, test/test_nn.py::TestNN::test_LayerNorm_3d_no_affine_large_feature_cuda, test/test_nn.py::TestNN::test_LayerNorm_3d_no_affine_large_feature_eval, test/test_nn.py::TestNN::test_LayerNorm_3d_no_affine_large_feature_eval_cuda, test/test_nn.py::TestNN::test_LeakyReLU_no_batch_dim, test/test_nn.py::TestNN::test_LeakyReLU_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Linear, test/test_nn.py::TestNN::test_Linear_cuda, test/test_nn.py::TestNN::test_Linear_no_batch_dim, test/test_nn.py::TestNN::test_Linear_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Linear_no_bias, test/test_nn.py::TestNN::test_Linear_no_bias_cuda, test/test_nn.py::TestNN::test_LogSigmoid_no_batch_dim, test/test_nn.py::TestNN::test_LogSigmoid_no_batch_dim_cuda, test/test_nn.py::TestNN::test_MSELoss_no_batch_dim_mean, test/test_nn.py::TestNN::test_MSELoss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_MSELoss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_MSELoss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_MSELoss_no_batch_dim_none, test/test_nn.py::TestNN::test_MSELoss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_MSELoss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_MSELoss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_MSELoss_no_batch_dim_sum, test/test_nn.py::TestNN::test_MSELoss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_MSELoss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_MSELoss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_MSELoss_no_reduce, test/test_nn.py::TestNN::test_MSELoss_no_reduce_cuda, test/test_nn.py::TestNN::test_MSELoss_no_reduce_scalar, test/test_nn.py::TestNN::test_MSELoss_no_reduce_scalar_cuda, test/test_nn.py::TestNN::test_MarginRankingLoss_no_batch_dim_mean, test/test_nn.py::TestNN::test_MarginRankingLoss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_MarginRankingLoss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_MarginRankingLoss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_MarginRankingLoss_no_batch_dim_none, test/test_nn.py::TestNN::test_MarginRankingLoss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_MarginRankingLoss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_MarginRankingLoss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_MarginRankingLoss_no_batch_dim_sum, test/test_nn.py::TestNN::test_MarginRankingLoss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_MarginRankingLoss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_MarginRankingLoss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_MaxUnpool1d_net, test/test_nn.py::TestNN::test_MaxUnpool1d_net_cuda, test/test_nn.py::TestNN::test_MaxUnpool1d_net_no_batch_dim, test/test_nn.py::TestNN::test_MaxUnpool1d_net_no_batch_dim_cuda, test/test_nn.py::TestNN::test_MaxUnpool2d_net, test/test_nn.py::TestNN::test_MaxUnpool2d_net_cuda, test/test_nn.py::TestNN::test_MaxUnpool2d_net_no_batch_dim, test/test_nn.py::TestNN::test_MaxUnpool2d_net_no_batch_dim_cuda, test/test_nn.py::TestNN::test_MaxUnpool3d_net, test/test_nn.py::TestNN::test_MaxUnpool3d_net_cuda, test/test_nn.py::TestNN::test_MaxUnpool3d_net_no_batch_dim, test/test_nn.py::TestNN::test_MaxUnpool3d_net_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Mish_no_batch_dim, test/test_nn.py::TestNN::test_Mish_no_batch_dim_cuda, test/test_nn.py::TestNN::test_ModuleDict, test/test_nn.py::TestNN::test_ModuleList, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_0d_no_reduce, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_0d_no_reduce_cuda, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_1d_no_reduce, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_1d_no_reduce_cuda, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_index_neg, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_index_neg_cuda, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_no_batch_dim_mean, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_no_batch_dim_none, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_no_batch_dim_sum, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_no_reduce, test/test_nn.py::TestNN::test_MultiLabelMarginLoss_no_reduce_cuda, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_no_batch_dim_mean, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_no_batch_dim_none, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_no_batch_dim_sum, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_no_reduce, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_no_reduce_cuda, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_weights_no_reduce, test/test_nn.py::TestNN::test_MultiLabelSoftMarginLoss_weights_no_reduce_cuda, test/test_nn.py::TestNN::test_MultiMarginLoss_1d_no_reduce, test/test_nn.py::TestNN::test_MultiMarginLoss_1d_no_reduce_cuda, test/test_nn.py::TestNN::test_MultiMarginLoss_margin_no_reduce, test/test_nn.py::TestNN::test_MultiMarginLoss_margin_no_reduce_cuda, test/test_nn.py::TestNN::test_MultiMarginLoss_no_reduce, test/test_nn.py::TestNN::test_MultiMarginLoss_no_reduce_cuda, test/test_nn.py::TestNN::test_MultiMarginLoss_p_no_reduce, test/test_nn.py::TestNN::test_MultiMarginLoss_p_no_reduce_cuda, test/test_nn.py::TestNN::test_MultiMarginLoss_weights_no_reduce, test/test_nn.py::TestNN::test_MultiMarginLoss_weights_no_reduce_cuda, test/test_nn.py::TestNN::test_NLLLoss2d_no_reduce, test/test_nn.py::TestNN::test_NLLLoss2d_no_reduce_cuda, test/test_nn.py::TestNN::test_NLLLoss2d_no_reduce_ignore_index, test/test_nn.py::TestNN::test_NLLLoss2d_no_reduce_ignore_index_cuda, test/test_nn.py::TestNN::test_NLLLoss2d_no_reduce_weights, test/test_nn.py::TestNN::test_NLLLoss2d_no_reduce_weights_cuda, test/test_nn.py::TestNN::test_NLLLossNd_no_reduce, test/test_nn.py::TestNN::test_NLLLossNd_no_reduce_cuda, test/test_nn.py::TestNN::test_NLLLossNd_no_reduce_ignore_index, test/test_nn.py::TestNN::test_NLLLossNd_no_reduce_ignore_index_cuda, test/test_nn.py::TestNN::test_NLLLossNd_no_reduce_weights, test/test_nn.py::TestNN::test_NLLLossNd_no_reduce_weights_cuda, test/test_nn.py::TestNN::test_NLLLoss_no_batch_dim_mean, test/test_nn.py::TestNN::test_NLLLoss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_NLLLoss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_NLLLoss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_NLLLoss_no_batch_dim_none, test/test_nn.py::TestNN::test_NLLLoss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_NLLLoss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_NLLLoss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_NLLLoss_no_batch_dim_sum, test/test_nn.py::TestNN::test_NLLLoss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_NLLLoss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_NLLLoss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_NLLLoss_no_reduce, test/test_nn.py::TestNN::test_NLLLoss_no_reduce_cuda, test/test_nn.py::TestNN::test_NLLLoss_no_reduce_ignore_index, test/test_nn.py::TestNN::test_NLLLoss_no_reduce_ignore_index_cuda, test/test_nn.py::TestNN::test_NLLLoss_no_reduce_weights, test/test_nn.py::TestNN::test_NLLLoss_no_reduce_weights_cuda, test/test_nn.py::TestNN::test_NLLLoss_no_reduce_weights_ignore_index, test/test_nn.py::TestNN::test_NLLLoss_no_reduce_weights_ignore_index_cuda, test/test_nn.py::TestNN::test_NLLLoss_no_reduce_weights_ignore_index_neg, test/test_nn.py::TestNN::test_NLLLoss_no_reduce_weights_ignore_index_neg_cuda, test/test_nn.py::TestNN::test_PReLU_backward_requires_grad_false, test/test_nn.py::TestNN::test_PReLU_no_batch_dim, test/test_nn.py::TestNN::test_PReLU_no_batch_dim_cuda, test/test_nn.py::TestNN::test_PairwiseDistance, test/test_nn.py::TestNN::test_PairwiseDistance_broadcast_lhs, test/test_nn.py::TestNN::test_PairwiseDistance_broadcast_lhs_cuda, test/test_nn.py::TestNN::test_PairwiseDistance_broadcast_rhs, test/test_nn.py::TestNN::test_PairwiseDistance_broadcast_rhs_cuda, test/test_nn.py::TestNN::test_PairwiseDistance_cuda, test/test_nn.py::TestNN::test_PairwiseDistance_no_batch_dim, test/test_nn.py::TestNN::test_PairwiseDistance_no_batch_dim_cuda, test/test_nn.py::TestNN::test_PairwiseDistance_with_non_default_args, test/test_nn.py::TestNN::test_PairwiseDistance_with_non_default_args_cuda, test/test_nn.py::TestNN::test_ParameterDict, test/test_nn.py::TestNN::test_ParameterDict_replication, test/test_nn.py::TestNN::test_ParameterList, test/test_nn.py::TestNN::test_ParameterList_meta, test/test_nn.py::TestNN::test_ParameterList_replication, test/test_nn.py::TestNN::test_PixelShuffle, test/test_nn.py::TestNN::test_PixelShuffle_cuda, test/test_nn.py::TestNN::test_PixelUnshuffle, test/test_nn.py::TestNN::test_PixelUnshuffle_cuda, test/test_nn.py::TestNN::test_PoissonNLLLoss_no_batch_dim_mean, test/test_nn.py::TestNN::test_PoissonNLLLoss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_PoissonNLLLoss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_PoissonNLLLoss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_PoissonNLLLoss_no_batch_dim_none, test/test_nn.py::TestNN::test_PoissonNLLLoss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_PoissonNLLLoss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_PoissonNLLLoss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_PoissonNLLLoss_no_batch_dim_sum, test/test_nn.py::TestNN::test_PoissonNLLLoss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_PoissonNLLLoss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_PoissonNLLLoss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_PoissonNLLLoss_no_reduce, test/test_nn.py::TestNN::test_PoissonNLLLoss_no_reduce_cuda, test/test_nn.py::TestNN::test_RNN_cell, test/test_nn.py::TestNN::test_RNN_cell_forward_zero_hidden_size, test/test_nn.py::TestNN::test_RNN_cell_no_broadcasting, test/test_nn.py::TestNN::test_RNN_change_dropout, test/test_nn.py::TestNN::test_RNN_cpu_vs_cudnn_no_dropout, test/test_nn.py::TestNN::test_RNN_cpu_vs_cudnn_with_dropout, test/test_nn.py::TestNN::test_RNN_cudnn_weight_norm, test/test_nn.py::TestNN::test_RNN_dropout, test/test_nn.py::TestNN::test_RNN_dropout_state, test/test_nn.py::TestNN::test_RNN_input_size_zero, test/test_nn.py::TestNN::test_RNN_nonlinearity, test/test_nn.py::TestNN::test_RNN_nonlinearity_passed_as_arg, test/test_nn.py::TestNN::test_RReLU, test/test_nn.py::TestNN::test_RReLU_cuda, test/test_nn.py::TestNN::test_RReLU_no_batch_dim, test/test_nn.py::TestNN::test_RReLU_no_batch_dim_cuda, test/test_nn.py::TestNN::test_RReLU_with_up_down, test/test_nn.py::TestNN::test_RReLU_with_up_down_cuda, test/test_nn.py::TestNN::test_RReLU_with_up_down_scalar, test/test_nn.py::TestNN::test_RReLU_with_up_down_scalar_cuda, test/test_nn.py::TestNN::test_ReLU6_no_batch_dim, test/test_nn.py::TestNN::test_ReLU6_no_batch_dim_cuda, test/test_nn.py::TestNN::test_ReLU_no_batch_dim, test/test_nn.py::TestNN::test_ReLU_no_batch_dim_cuda, test/test_nn.py::TestNN::test_ReplicationPad3d, test/test_nn.py::TestNN::test_ReplicationPad3d_complex, test/test_nn.py::TestNN::test_ReplicationPad3d_complex_cuda, test/test_nn.py::TestNN::test_ReplicationPad3d_cuda, test/test_nn.py::TestNN::test_ReplicationPad3d_no_batch_dim, test/test_nn.py::TestNN::test_ReplicationPad3d_no_batch_dim_cuda, test/test_nn.py::TestNN::test_SELU_no_batch_dim, test/test_nn.py::TestNN::test_SELU_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Sequential_add, test/test_nn.py::TestNN::test_Sequential_append, test/test_nn.py::TestNN::test_Sequential_delitem, test/test_nn.py::TestNN::test_Sequential_extend, test/test_nn.py::TestNN::test_Sequential_getitem, test/test_nn.py::TestNN::test_Sequential_iadd, test/test_nn.py::TestNN::test_Sequential_imul, test/test_nn.py::TestNN::test_Sequential_insert, test/test_nn.py::TestNN::test_Sequential_insert_fail_case, test/test_nn.py::TestNN::test_Sequential_mul, test/test_nn.py::TestNN::test_Sequential_pop, test/test_nn.py::TestNN::test_Sequential_rmul, test/test_nn.py::TestNN::test_Sequential_setitem, test/test_nn.py::TestNN::test_Sequential_setitem_named, test/test_nn.py::TestNN::test_SiLU_no_batch_dim, test/test_nn.py::TestNN::test_SiLU_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Sigmoid_no_batch_dim, test/test_nn.py::TestNN::test_Sigmoid_no_batch_dim_cuda, test/test_nn.py::TestNN::test_SmoothL1Loss_beta, test/test_nn.py::TestNN::test_SmoothL1Loss_beta_cuda, test/test_nn.py::TestNN::test_SmoothL1Loss_no_batch_dim_mean, test/test_nn.py::TestNN::test_SmoothL1Loss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_SmoothL1Loss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_SmoothL1Loss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_SmoothL1Loss_no_batch_dim_none, test/test_nn.py::TestNN::test_SmoothL1Loss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_SmoothL1Loss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_SmoothL1Loss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_SmoothL1Loss_no_batch_dim_sum, test/test_nn.py::TestNN::test_SmoothL1Loss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_SmoothL1Loss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_SmoothL1Loss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_SmoothL1Loss_no_reduce, test/test_nn.py::TestNN::test_SmoothL1Loss_no_reduce_cuda, test/test_nn.py::TestNN::test_SmoothL1Loss_no_reduce_scalar, test/test_nn.py::TestNN::test_SmoothL1Loss_no_reduce_scalar_cuda, test/test_nn.py::TestNN::test_SmoothL1Loss_zero_beta, test/test_nn.py::TestNN::test_SmoothL1Loss_zero_beta_cuda, test/test_nn.py::TestNN::test_SoftMarginLoss_no_batch_dim_mean, test/test_nn.py::TestNN::test_SoftMarginLoss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_SoftMarginLoss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_SoftMarginLoss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_SoftMarginLoss_no_batch_dim_none, test/test_nn.py::TestNN::test_SoftMarginLoss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_SoftMarginLoss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_SoftMarginLoss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_SoftMarginLoss_no_batch_dim_sum, test/test_nn.py::TestNN::test_SoftMarginLoss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_SoftMarginLoss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_SoftMarginLoss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_SoftMarginLoss_no_reduce, test/test_nn.py::TestNN::test_SoftMarginLoss_no_reduce_cuda, test/test_nn.py::TestNN::test_Softplus_no_batch_dim, test/test_nn.py::TestNN::test_Softplus_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Softshrink_no_batch_dim, test/test_nn.py::TestNN::test_Softshrink_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Softsign_no_batch_dim, test/test_nn.py::TestNN::test_Softsign_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Tanh_no_batch_dim, test/test_nn.py::TestNN::test_Tanh_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Tanhshrink_no_batch_dim, test/test_nn.py::TestNN::test_Tanhshrink_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Threshold_no_batch_dim, test/test_nn.py::TestNN::test_Threshold_no_batch_dim_cuda, test/test_nn.py::TestNN::test_TransformerDecoderLayer_gelu_activation, test/test_nn.py::TestNN::test_TransformerDecoderLayer_gelu_activation_cuda, test/test_nn.py::TestNN::test_TransformerDecoderLayer_relu_activation, test/test_nn.py::TestNN::test_TransformerDecoderLayer_relu_activation_cuda, test/test_nn.py::TestNN::test_TransformerEncoderLayer_gelu_activation, test/test_nn.py::TestNN::test_TransformerEncoderLayer_gelu_activation_cuda, test/test_nn.py::TestNN::test_TransformerEncoderLayer_relu_activation, test/test_nn.py::TestNN::test_TransformerEncoderLayer_relu_activation_cuda, test/test_nn.py::TestNN::test_Transformer_cell, test/test_nn.py::TestNN::test_Transformer_multilayer_coder, test/test_nn.py::TestNN::test_Transformer_multilayer_coder_cuda, test/test_nn.py::TestNN::test_TripletMarginLoss_no_batch_dim_mean, test/test_nn.py::TestNN::test_TripletMarginLoss_no_batch_dim_mean_cuda_double, test/test_nn.py::TestNN::test_TripletMarginLoss_no_batch_dim_mean_cuda_float, test/test_nn.py::TestNN::test_TripletMarginLoss_no_batch_dim_mean_cuda_half, test/test_nn.py::TestNN::test_TripletMarginLoss_no_batch_dim_none, test/test_nn.py::TestNN::test_TripletMarginLoss_no_batch_dim_none_cuda_double, test/test_nn.py::TestNN::test_TripletMarginLoss_no_batch_dim_none_cuda_float, test/test_nn.py::TestNN::test_TripletMarginLoss_no_batch_dim_none_cuda_half, test/test_nn.py::TestNN::test_TripletMarginLoss_no_batch_dim_sum, test/test_nn.py::TestNN::test_TripletMarginLoss_no_batch_dim_sum_cuda_double, test/test_nn.py::TestNN::test_TripletMarginLoss_no_batch_dim_sum_cuda_float, test/test_nn.py::TestNN::test_TripletMarginLoss_no_batch_dim_sum_cuda_half, test/test_nn.py::TestNN::test_Unflatten_no_batch_dim, test/test_nn.py::TestNN::test_Unflatten_no_batch_dim_cuda, test/test_nn.py::TestNN::test_Unfold, test/test_nn.py::TestNN::test_Unfold_cuda, test/test_nn.py::TestNN::test_Unfold_int_input, test/test_nn.py::TestNN::test_Unfold_int_input_cuda, test/test_nn.py::TestNN::test_adaptive_log_softmax, test/test_nn.py::TestNN::test_add_module, test/test_nn.py::TestNN::test_add_module_raises_error_if_attr_exists, test/test_nn.py::TestNN::test_affine_grid, test/test_nn.py::TestNN::test_affine_grid_3d, test/test_nn.py::TestNN::test_affine_grid_backward_cl_cf_consistency_device_cpu_nd_2, test/test_nn.py::TestNN::test_affine_grid_backward_cl_cf_consistency_device_cpu_nd_3, test/test_nn.py::TestNN::test_affine_grid_backward_cl_cf_consistency_device_cuda_nd_2, test/test_nn.py::TestNN::test_affine_grid_backward_cl_cf_consistency_device_cuda_nd_3, test/test_nn.py::TestNN::test_affine_grid_error_checking, test/test_nn.py::TestNN::test_assignment, test/test_nn.py::TestNN::test_batch_norm_update_stats, test/test_nn.py::TestNN::test_batchnorm_buffer_update_when_stats_are_not_tracked, test/test_nn.py::TestNN::test_batchnorm_cudnn_half, test/test_nn.py::TestNN::test_batchnorm_cudnn_nhwc, test/test_nn.py::TestNN::test_batchnorm_load_state_dict, test/test_nn.py::TestNN::test_batchnorm_nhwc_cpu, test/test_nn.py::TestNN::test_batchnorm_nhwc_cuda, test/test_nn.py::TestNN::test_batchnorm_non_contig_cpu_BatchNorm2d, test/test_nn.py::TestNN::test_batchnorm_non_contig_cpu_SyncBatchNorm, test/test_nn.py::TestNN::test_batchnorm_nonaffine_cuda_half_input, test/test_nn.py::TestNN::test_batchnorm_raises_error_if_bias_is_not_same_size_as_input, test/test_nn.py::TestNN::test_batchnorm_raises_error_if_less_than_one_value_per_channel, test/test_nn.py::TestNN::test_batchnorm_raises_error_if_running_mean_is_not_same_size_as_input, test/test_nn.py::TestNN::test_batchnorm_raises_error_if_running_var_is_not_same_size_as_input, test/test_nn.py::TestNN::test_batchnorm_raises_error_if_running_var_or_running_mean_have_forward_grad, test/test_nn.py::TestNN::test_batchnorm_raises_error_if_weight_is_not_same_size_as_input, test/test_nn.py::TestNN::test_bce_loss_always_nonnegative, test/test_nn.py::TestNN::test_bce_loss_broadcasts_weights, test/test_nn.py::TestNN::test_bce_loss_input_range, test/test_nn.py::TestNN::test_bce_loss_size_mismatch, test/test_nn.py::TestNN::test_bce_with_logits_broadcasts_pos_weights, test/test_nn.py::TestNN::test_bce_with_logits_broadcasts_weights, test/test_nn.py::TestNN::test_bce_with_logits_gives_same_result_as_sigmoid_and_bce_loss, test/test_nn.py::TestNN::test_bce_with_logits_gives_same_result_as_sigmoid_and_bce_loss_large_tensors_with_grad, test/test_nn.py::TestNN::test_bce_with_logits_has_correct_forward_grad, test/test_nn.py::TestNN::test_bce_with_logits_has_correct_grad_at_zero, test/test_nn.py::TestNN::test_bce_with_logits_ones_in_pos_weights_are_the_same_as_none, test/test_nn.py::TestNN::test_bce_with_logits_raises_if_target_and_input_are_different_size, test/test_nn.py::TestNN::test_bce_with_logits_stability, test/test_nn.py::TestNN::test_bce_with_logits_with_pos_weight_has_correct_grad_at_zero, test/test_nn.py::TestNN::test_bilinear, test/test_nn.py::TestNN::test_bilinear_broadcasting, test/test_nn.py::TestNN::test_bilinear_no_bias, test/test_nn.py::TestNN::test_bilinear_non_contiguous, test/test_nn.py::TestNN::test_broadcast_double_backwards_gpu, test/test_nn.py::TestNN::test_broadcast_no_grad, test/test_nn.py::TestNN::test_broadcast_not_requiring_grad, test/test_nn.py::TestNN::test_buffer_bad_module_subclass, test/test_nn.py::TestNN::test_buffer_not_persistent, test/test_nn.py::TestNN::test_buffer_not_persistent_assign, test/test_nn.py::TestNN::test_buffer_not_persistent_del, test/test_nn.py::TestNN::test_buffer_not_persistent_load, test/test_nn.py::TestNN::test_buffer_not_persistent_overwrite, test/test_nn.py::TestNN::test_buffers_and_named_buffers, test/test_nn.py::TestNN::test_call_supports_python_dict_output, test/test_nn.py::TestNN::test_channel_shuffle_return_alias_of_self, test/test_nn.py::TestNN::test_children, test/test_nn.py::TestNN::test_container_copy, test/test_nn.py::TestNN::test_convert_sync_batchnorm, test/test_nn.py::TestNN::test_cosine_embedding_loss_error_on_diff_shapes, test/test_nn.py::TestNN::test_cosine_embedding_loss_error_on_nonexpandable_shapes, test/test_nn.py::TestNN::test_cosine_embedding_loss_invalid_shape, test/test_nn.py::TestNN::test_cosine_embedding_loss_margin_no_reduce, test/test_nn.py::TestNN::test_cosine_embedding_loss_no_reduce, test/test_nn.py::TestNN::test_cosine_embedding_loss_with_diff_type, test/test_nn.py::TestNN::test_cosine_similarity, test/test_nn.py::TestNN::test_cross_entropy_loss, test/test_nn.py::TestNN::test_cross_entropy_loss_precision, test/test_nn.py::TestNN::test_cross_entropy_loss_zero_div, test/test_nn.py::TestNN::test_cudnn_forward_exception, test/test_nn.py::TestNN::test_cudnn_rnn_dropout_states_device, test/test_nn.py::TestNN::test_cudnn_weight_format, test/test_nn.py::TestNN::test_cudnn_weight_tying, test/test_nn.py::TestNN::test_dir, test/test_nn.py::TestNN::test_dir_digit, test/test_nn.py::TestNN::test_elu_inplace_gradgrad, test/test_nn.py::TestNN::test_elu_inplace_on_view, test/test_nn.py::TestNN::test_error_RNN_seq_len_zero, test/test_nn.py::TestNN::test_extra_state, test/test_nn.py::TestNN::test_extra_state_missing_get_extra_state, test/test_nn.py::TestNN::test_extra_state_missing_set_extra_state, test/test_nn.py::TestNN::test_extra_state_non_dict, test/test_nn.py::TestNN::test_fb_fc_packed, test/test_nn.py::TestNN::test_flatten, test/test_nn.py::TestNN::test_fold_invalid_arg, test/test_nn.py::TestNN::test_fractional_max_pool2d_invalid_output_ratio, test/test_nn.py::TestNN::test_gaussian_nll_loss_args, test/test_nn.py::TestNN::test_gaussian_nll_loss_broadcasting, test/test_nn.py::TestNN::test_get_buffer, test/test_nn.py::TestNN::test_get_buffer_from_submodules, test/test_nn.py::TestNN::test_getattr_with_property, test/test_nn.py::TestNN::test_grid_sample, test/test_nn.py::TestNN::test_grid_sample_3d, test/test_nn.py::TestNN::test_grid_sample_error_checking, test/test_nn.py::TestNN::test_grid_sample_nearest_neighbor_rounding_mode_consistency, test/test_nn.py::TestNN::test_hardtanh_backward, test/test_nn.py::TestNN::test_hardtanh_inplace_gradgrad, test/test_nn.py::TestNN::test_huber_loss_invalid_delta, test/test_nn.py::TestNN::test_inplace_thnn, test/test_nn.py::TestNN::test_interpolate, test/test_nn.py::TestNN::test_interpolate_bicubic_2d, test/test_nn.py::TestNN::test_interpolate_bicubic_2d_cuda, test/test_nn.py::TestNN::test_interpolate_bicubic_2d_zero_dim, test/test_nn.py::TestNN::test_interpolate_bicubic_2d_zero_dim_cuda, test/test_nn.py::TestNN::test_interpolate_bicubic_scale_2d, test/test_nn.py::TestNN::test_interpolate_bicubic_scale_2d_cuda, test/test_nn.py::TestNN::test_interpolate_bicubic_scale_tuple_shared_2d, test/test_nn.py::TestNN::test_interpolate_bicubic_scale_tuple_shared_2d_cuda, test/test_nn.py::TestNN::test_interpolate_bicubic_scale_tuple_skewed_2d, test/test_nn.py::TestNN::test_interpolate_bicubic_scale_tuple_skewed_2d_align_corners, test/test_nn.py::TestNN::test_interpolate_bicubic_scale_tuple_skewed_2d_align_corners_cuda, test/test_nn.py::TestNN::test_interpolate_bicubic_scale_tuple_skewed_2d_cuda, test/test_nn.py::TestNN::test_interpolate_bicubic_tuple_2d, test/test_nn.py::TestNN::test_interpolate_bicubic_tuple_2d_align_corners, test/test_nn.py::TestNN::test_interpolate_bicubic_tuple_2d_align_corners_cuda, test/test_nn.py::TestNN::test_interpolate_bicubic_tuple_2d_cuda, test/test_nn.py::TestNN::test_interpolate_bilinear_2d, test/test_nn.py::TestNN::test_interpolate_bilinear_2d_cuda, test/test_nn.py::TestNN::test_interpolate_bilinear_2d_zero_dim, test/test_nn.py::TestNN::test_interpolate_bilinear_2d_zero_dim_cuda, test/test_nn.py::TestNN::test_interpolate_bilinear_scale_2d, test/test_nn.py::TestNN::test_interpolate_bilinear_scale_2d_cuda, test/test_nn.py::TestNN::test_interpolate_bilinear_scale_tuple_shared_2d, test/test_nn.py::TestNN::test_interpolate_bilinear_scale_tuple_shared_2d_cuda, test/test_nn.py::TestNN::test_interpolate_bilinear_scale_tuple_skewed_2d, test/test_nn.py::TestNN::test_interpolate_bilinear_scale_tuple_skewed_2d_align_corners, test/test_nn.py::TestNN::test_interpolate_bilinear_scale_tuple_skewed_2d_align_corners_cuda, test/test_nn.py::TestNN::test_interpolate_bilinear_scale_tuple_skewed_2d_cuda, test/test_nn.py::TestNN::test_interpolate_bilinear_tuple_2d, test/test_nn.py::TestNN::test_interpolate_bilinear_tuple_2d_align_corners, test/test_nn.py::TestNN::test_interpolate_bilinear_tuple_2d_align_corners_cuda, test/test_nn.py::TestNN::test_interpolate_bilinear_tuple_2d_cuda, test/test_nn.py::TestNN::test_interpolate_buffer_overflow, test/test_nn.py::TestNN::test_interpolate_illegal_memory_access, test/test_nn.py::TestNN::test_interpolate_linear_1d, test/test_nn.py::TestNN::test_interpolate_linear_1d_align_corners, test/test_nn.py::TestNN::test_interpolate_linear_1d_align_corners_cuda, test/test_nn.py::TestNN::test_interpolate_linear_1d_cuda, test/test_nn.py::TestNN::test_interpolate_linear_1d_zero_dim, test/test_nn.py::TestNN::test_interpolate_linear_1d_zero_dim_cuda, test/test_nn.py::TestNN::test_interpolate_linear_scale_1d, test/test_nn.py::TestNN::test_interpolate_linear_scale_1d_align_corners, test/test_nn.py::TestNN::test_interpolate_linear_scale_1d_align_corners_cuda, test/test_nn.py::TestNN::test_interpolate_linear_scale_1d_cuda, test/test_nn.py::TestNN::test_interpolate_linear_tuple_1d, test/test_nn.py::TestNN::test_interpolate_linear_tuple_1d_cuda, test/test_nn.py::TestNN::test_interpolate_nearest_1d, test/test_nn.py::TestNN::test_interpolate_nearest_1d_cuda, test/test_nn.py::TestNN::test_interpolate_nearest_1d_zero_dim, test/test_nn.py::TestNN::test_interpolate_nearest_1d_zero_dim_cuda, test/test_nn.py::TestNN::test_interpolate_nearest_2d, test/test_nn.py::TestNN::test_interpolate_nearest_2d_cuda, test/test_nn.py::TestNN::test_interpolate_nearest_2d_launch_configs, test/test_nn.py::TestNN::test_interpolate_nearest_2d_launch_configs_cuda, test/test_nn.py::TestNN::test_interpolate_nearest_2d_zero_dim, test/test_nn.py::TestNN::test_interpolate_nearest_2d_zero_dim_cuda, test/test_nn.py::TestNN::test_interpolate_nearest_3d, test/test_nn.py::TestNN::test_interpolate_nearest_3d_cuda, test/test_nn.py::TestNN::test_interpolate_nearest_3d_zero_dim, test/test_nn.py::TestNN::test_interpolate_nearest_3d_zero_dim_cuda, test/test_nn.py::TestNN::test_interpolate_nearest_scale_1d, test/test_nn.py::TestNN::test_interpolate_nearest_scale_1d_cuda, test/test_nn.py::TestNN::test_interpolate_nearest_scale_2d, test/test_nn.py::TestNN::test_interpolate_nearest_scale_2d_cuda, test/test_nn.py::TestNN::test_interpolate_nearest_scale_3d, test/test_nn.py::TestNN::test_interpolate_nearest_scale_3d_cuda, test/test_nn.py::TestNN::test_interpolate_nearest_tuple_1d, test/test_nn.py::TestNN::test_interpolate_nearest_tuple_1d_cuda, test/test_nn.py::TestNN::test_interpolate_nearest_tuple_2d, test/test_nn.py::TestNN::test_interpolate_nearest_tuple_2d_cuda, test/test_nn.py::TestNN::test_interpolate_nearest_tuple_3d, test/test_nn.py::TestNN::test_interpolate_nearest_tuple_3d_cuda, test/test_nn.py::TestNN::test_interpolate_trilinear_3d, test/test_nn.py::TestNN::test_interpolate_trilinear_3d_cuda, test/test_nn.py::TestNN::test_interpolate_trilinear_3d_zero_dim, test/test_nn.py::TestNN::test_interpolate_trilinear_3d_zero_dim_cuda, test/test_nn.py::TestNN::test_interpolate_trilinear_scale_3d, test/test_nn.py::TestNN::test_interpolate_trilinear_scale_3d_align_corners, test/test_nn.py::TestNN::test_interpolate_trilinear_scale_3d_align_corners_cuda, test/test_nn.py::TestNN::test_interpolate_trilinear_scale_3d_cuda, test/test_nn.py::TestNN::test_interpolate_trilinear_tuple_3d, test/test_nn.py::TestNN::test_interpolate_trilinear_tuple_3d_align_corners, test/test_nn.py::TestNN::test_interpolate_trilinear_tuple_3d_align_corners_cuda, test/test_nn.py::TestNN::test_interpolate_trilinear_tuple_3d_cuda, test/test_nn.py::TestNN::test_interpolate_undefined_behavior_casting, test/test_nn.py::TestNN::test_kl_div_log_softmax_target, test/test_nn.py::TestNN::test_kl_div_with_diff_type, test/test_nn.py::TestNN::test_kl_div_with_diff_type_log_target, test/test_nn.py::TestNN::test_l1_loss_correct, test/test_nn.py::TestNN::test_layer_norm_eps, test/test_nn.py::TestNN::test_layer_norm_grads_with_create_graph_flag, test/test_nn.py::TestNN::test_linear_autograd_device_cpu_bias_weightCOO, test/test_nn.py::TestNN::test_linear_autograd_device_cpu_bias_weightCSC, test/test_nn.py::TestNN::test_linear_autograd_device_cpu_bias_weightCSR, test/test_nn.py::TestNN::test_linear_autograd_device_cpu_bias_weightStrided, test/test_nn.py::TestNN::test_linear_autograd_device_cpu_nobias_weightCOO, test/test_nn.py::TestNN::test_linear_autograd_device_cpu_nobias_weightCSC, test/test_nn.py::TestNN::test_linear_autograd_device_cpu_nobias_weightCSR, test/test_nn.py::TestNN::test_linear_autograd_device_cpu_nobias_weightStrided, test/test_nn.py::TestNN::test_linear_autograd_device_cuda_bias_weightCOO, test/test_nn.py::TestNN::test_linear_autograd_device_cuda_bias_weightCSC, test/test_nn.py::TestNN::test_linear_autograd_device_cuda_bias_weightCSR, test/test_nn.py::TestNN::test_linear_autograd_device_cuda_bias_weightStrided, test/test_nn.py::TestNN::test_linear_autograd_device_cuda_nobias_weightCOO, test/test_nn.py::TestNN::test_linear_autograd_device_cuda_nobias_weightCSC, test/test_nn.py::TestNN::test_linear_autograd_device_cuda_nobias_weightCSR, test/test_nn.py::TestNN::test_linear_autograd_device_cuda_nobias_weightStrided, test/test_nn.py::TestNN::test_linear_broadcasting, test/test_nn.py::TestNN::test_linear_raise_on_scalar_input, test/test_nn.py::TestNN::test_log_softmax_dim0, test/test_nn.py::TestNN::test_log_softmax_dim0_cuda, test/test_nn.py::TestNN::test_log_softmax_dim3, test/test_nn.py::TestNN::test_log_softmax_dim3_cuda, test/test_nn.py::TestNN::test_log_softmax_lastdim, test/test_nn.py::TestNN::test_log_softmax_lastdim_cuda, test/test_nn.py::TestNN::test_log_softmax_scalar, test/test_nn.py::TestNN::test_log_softmax_scalar_cuda, test/test_nn.py::TestNN::test_log_softmax_spatial, test/test_nn.py::TestNN::test_log_softmax_spatial_cuda, test/test_nn.py::TestNN::test_log_softmax_spatial_special, test/test_nn.py::TestNN::test_log_softmax_spatial_special_cuda, test/test_nn.py::TestNN::test_loss_equal_input_target_shape, test/test_nn.py::TestNN::test_margin_ranking_loss_margin_no_reduce, test/test_nn.py::TestNN::test_margin_ranking_loss_no_reduce, test/test_nn.py::TestNN::test_max_pool1d_invalid_output_size, test/test_nn.py::TestNN::test_module_apply_inplace_op, test/test_nn.py::TestNN::test_module_backcompat, test/test_nn.py::TestNN::test_module_super_init, test/test_nn.py::TestNN::test_module_to_argparse, test/test_nn.py::TestNN::test_modules, test/test_nn.py::TestNN::test_mse_loss_size_warning, test/test_nn.py::TestNN::test_multimarginloss_1d_input_0d_target_no_reduce, test/test_nn.py::TestNN::test_multimarginloss_1d_input_0d_target_no_reduce_cuda, test/test_nn.py::TestNN::test_named_children, test/test_nn.py::TestNN::test_named_modules, test/test_nn.py::TestNN::test_named_parameters_remove_duplicate, test/test_nn.py::TestNN::test_native_channel_shuffle_return_alias_of_self, test/test_nn.py::TestNN::test_nested_tensor_from_mask, test/test_nn.py::TestNN::test_nested_tensor_from_mask_error, test/test_nn.py::TestNN::test_no_grad, test/test_nn.py::TestNN::test_non_leaf_parameters, test/test_nn.py::TestNN::test_normalize, test/test_nn.py::TestNN::test_overwrite_module_params_on_conversion, test/test_nn.py::TestNN::test_pack_sequence_batch_sizes_throw, test/test_nn.py::TestNN::test_pad_scalar_error, test/test_nn.py::TestNN::test_padding_list, test/test_nn.py::TestNN::test_pairwise_distance, test/test_nn.py::TestNN::test_parameter_assignment, test/test_nn.py::TestNN::test_parameterlistdict_pickle, test/test_nn.py::TestNN::test_parameterlistdict_setting_attributes, test/test_nn.py::TestNN::test_parameters_and_named_parameters, test/test_nn.py::TestNN::test_parameters_to_vector, test/test_nn.py::TestNN::test_parse_to, test/test_nn.py::TestNN::test_partial_flat_weights, test/test_nn.py::TestNN::test_pdist, test/test_nn.py::TestNN::test_pdist_cpu_gradgrad_unimplemented, test/test_nn.py::TestNN::test_pdist_cuda_gradgrad_unimplemented, test/test_nn.py::TestNN::test_pdist_empty_col, test/test_nn.py::TestNN::test_pdist_empty_row, test/test_nn.py::TestNN::test_pdist_large, test/test_nn.py::TestNN::test_pdist_zeros, test/test_nn.py::TestNN::test_pickle_module_no_weights_only_warning, test/test_nn.py::TestNN::test_pixel_shuffle_nhwc_cpu, test/test_nn.py::TestNN::test_pixel_shuffle_unshuffle, test/test_nn.py::TestNN::test_pointwise_loss_broadcast, test/test_nn.py::TestNN::test_pointwise_loss_target_grad_none_reduction, test/test_nn.py::TestNN::test_projections_errors_on_gru_and_rnn, test/test_nn.py::TestNN::test_projections_lstm_args_check, test/test_nn.py::TestNN::test_projections_lstm_check_device, test/test_nn.py::TestNN::test_projections_lstm_initial_hidden_state, test/test_nn.py::TestNN::test_register_buffer_allows_overwriting_with_same_name, test/test_nn.py::TestNN::test_register_buffer_raises_error_if_attr_exists, test/test_nn.py::TestNN::test_register_buffer_raises_error_if_name_is_not_string, test/test_nn.py::TestNN::test_register_buffer_raises_error_if_not_tensor, test/test_nn.py::TestNN::test_register_parameter_allows_overwriting_with_same_name, test/test_nn.py::TestNN::test_register_parameter_raises_error_if_attr_exists, test/test_nn.py::TestNN::test_register_parameter_raises_error_if_name_is_not_string, test/test_nn.py::TestNN::test_relu_inplace_on_view, test/test_nn.py::TestNN::test_repr, test/test_nn.py::TestNN::test_requires_grad_, test/test_nn.py::TestNN::test_rnn_args_check, test/test_nn.py::TestNN::test_rnn_check_device, test/test_nn.py::TestNN::test_rnn_initial_hidden_state, test/test_nn.py::TestNN::test_rnn_weight_norm, test/test_nn.py::TestNN::test_set_submodule, test/test_nn.py::TestNN::test_share_memory, test/test_nn.py::TestNN::test_smoothl1loss_intergral_target, test/test_nn.py::TestNN::test_smoothl1loss_negative_beta_not_supported, test/test_nn.py::TestNN::test_softmax_functional_dim0, test/test_nn.py::TestNN::test_softmax_functional_dim0_cuda, test/test_nn.py::TestNN::test_softmax_functional_dim3, test/test_nn.py::TestNN::test_softmax_functional_dim3_cuda, test/test_nn.py::TestNN::test_softmax_functional_scalar, test/test_nn.py::TestNN::test_softmax_functional_scalar_cuda, test/test_nn.py::TestNN::test_softmax_lastdim, test/test_nn.py::TestNN::test_softmax_lastdim_cuda, test/test_nn.py::TestNN::test_softmax_lastdim_dtype, test/test_nn.py::TestNN::test_softmax_lastdim_dtype_cuda, test/test_nn.py::TestNN::test_softmax_spatial, test/test_nn.py::TestNN::test_softmax_spatial_cuda, test/test_nn.py::TestNN::test_softmax_spatial_dtype, test/test_nn.py::TestNN::test_softmax_spatial_dtype_cuda, test/test_nn.py::TestNN::test_softmax_spatial_special, test/test_nn.py::TestNN::test_softmax_spatial_special_cuda, test/test_nn.py::TestNN::test_softmin, test/test_nn.py::TestNN::test_spectral_norm, test/test_nn.py::TestNN::test_spectral_norm_dim, test/test_nn.py::TestNN::test_spectral_norm_forward, test/test_nn.py::TestNN::test_spectral_norm_load_state_dict, test/test_nn.py::TestNN::test_spectral_norm_pickle, test/test_nn.py::TestNN::test_state_dict, test/test_nn.py::TestNN::test_swap_module_params_poisons_acc_grad, test/test_nn.py::TestNN::test_sync_batchnorm_accuracy_cuda, test/test_nn.py::TestNN::test_sync_batchnorm_backward_elemt, test/test_nn.py::TestNN::test_threshold_bfloat16_half, test/test_nn.py::TestNN::test_threshold_int, test/test_nn.py::TestNN::test_to, test/test_nn.py::TestNN::test_train_errors_for_invalid_mode, test/test_nn.py::TestNN::test_transformer_args_check, test/test_nn.py::TestNN::test_transformer_layer_args_check, test/test_nn.py::TestNN::test_transformerdecoder, test/test_nn.py::TestNN::test_transformerdecoderlayer, test/test_nn.py::TestNN::test_transformerdecoderlayer_gelu, test/test_nn.py::TestNN::test_triplet_margin_loss, test/test_nn.py::TestNN::test_triplet_margin_loss_no_reduce, test/test_nn.py::TestNN::test_triplet_margin_loss_swap, test/test_nn.py::TestNN::test_triplet_margin_loss_swap_no_reduce, test/test_nn.py::TestNN::test_type, test/test_nn.py::TestNN::test_unflatten, test/test_nn.py::TestNN::test_unflatten_invalid_arg, test/test_nn.py::TestNN::test_unfold_invalid_arg, test/test_nn.py::TestNN::test_upsamplingBilinear2d_spatial_invariance, test/test_nn.py::TestNN::test_upsamplingLinear1d, test/test_nn.py::TestNN::test_upsamplingLinear1d_spatial_invariance, test/test_nn.py::TestNN::test_upsamplingTrilinear3d_spatial_invariance, test/test_nn.py::TestNN::test_upsampling_bfloat16, test/test_nn.py::TestNN::test_upsampling_not_recompute_scale_factor, test/test_nn.py::TestNN::test_upsampling_small_scale, test/test_nn.py::TestNN::test_vector_to_parameters, test/test_nn.py::TestNN::test_weight_norm, test/test_nn.py::TestNN::test_weight_norm_pickle, test/test_nn.py::TestNN::test_zero_grad, test/test_nn.py::TestFusionEval::test_fuse_module_eval_numerics, test/test_nn.py::TestConstantPadNd::test_constant_pad_nd, test/test_nn.py::TestConstantPadNd::test_preserves_memory_format, test/test_nn.py::TestAddRelu::test_add_relu, test/test_nn.py::TestAddRelu::test_add_relu_broadcasting, test/test_nn.py::TestFunctionalPickle::test_pickle_softsign, test/test_nn.py::TestFusionUtils::test_fuse_conv_bn_requires_grad, test/test_nn.py::TestFusionUtils::test_fuse_linear_bn_requires_grad, test/test_nn.py::TestUtils::test_consume_prefix_in_state_dict_if_present, test/test_nn.py::TestNNDeviceTypeCUDA::test_BatchNorm_empty_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_Bilinear_empty_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_CTCLoss_cudnn_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_CTCLoss_empty_target_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_CTCLoss_no_batch_dim_reduction_mean_use_module_form_False_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_CTCLoss_no_batch_dim_reduction_mean_use_module_form_True_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_CTCLoss_no_batch_dim_reduction_none_use_module_form_False_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_CTCLoss_no_batch_dim_reduction_none_use_module_form_True_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_CTCLoss_no_batch_dim_reduction_sum_use_module_form_False_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_CTCLoss_no_batch_dim_reduction_sum_use_module_form_True_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_GRU_grad_and_gradgrad_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_GroupNorm_empty_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_GroupNorm_general_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_GroupNorm_memory_format_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_GroupNorm_numeric_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_GroupNorm_raises_error_if_one_value_per_group_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_InstanceNorm1d_general_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_InstanceNorm2d_general_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_InstanceNorm3d_general_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_LSTM_differentiable_backward_using_oneDNN_cuda_bfloat16, test/test_nn.py::TestNNDeviceTypeCUDA::test_LSTM_differentiable_backward_using_oneDNN_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_LSTM_grad_and_gradgrad_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_LayerNorm_general_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_LayerNorm_numeric_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_LocalResponseNorm_empty_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_MarginLoss_empty_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_MarginLoss_empty_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_MarginLoss_warnings_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_ReflectionPad2d_large_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_ReflectionPad3d_large_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_ReflectionPad_empty_cuda_complex64, test/test_nn.py::TestNNDeviceTypeCUDA::test_ReflectionPad_empty_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_ReplicationPad1d_large_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_ReplicationPad2d_large_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_ReplicationPad3d_large_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_ReplicationPad_empty_cuda_complex128, test/test_nn.py::TestNNDeviceTypeCUDA::test_ReplicationPad_empty_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_TransformerDecoderLayer_empty_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_TransformerDecoder_empty_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_TransformerEncoderLayer_empty_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_TransformerEncoder_empty_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_Transformer_empty_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_Unfold_empty_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_activations_bfloat16_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_activations_bfloat16_half_cpu_cuda_bfloat16, test/test_nn.py::TestNNDeviceTypeCUDA::test_activations_bfloat16_half_cpu_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_adaptiveavg_pool1d_shmem_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_affine_2d_rotate0_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_affine_2d_rotate45_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_affine_2d_rotate90_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_affine_2d_rotateRandom_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_affine_3d_rotateRandom_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_avg_pool_large_tensor2_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_avg_pool_large_tensor_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_affine_cuda_bfloat16, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_affine_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_affine_mixed_cuda_bfloat16, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_affine_mixed_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_eval_cuda_bfloat16, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_eval_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_eval_mixed_cuda_bfloat16, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_eval_mixed_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_grad_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_large_batch_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_large_batch_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_simple_average_cuda_bfloat16, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_simple_average_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_simple_average_mixed_cuda_bfloat16, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_simple_average_mixed_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_batchnorm_update_stats_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_channel_shuffle_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_norm_error_if_nonfinite_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_norm_foreach_False_norm_type_0_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_norm_foreach_False_norm_type_1_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_norm_foreach_False_norm_type_2_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_norm_foreach_False_norm_type_4_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_norm_foreach_False_norm_type_inf_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_norm_foreach_True_norm_type_0_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_norm_foreach_True_norm_type_1_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_norm_foreach_True_norm_type_2_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_norm_foreach_True_norm_type_4_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_norm_foreach_True_norm_type_inf_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_norm_multi_device_foreach_False_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_norm_multi_device_foreach_True_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_value_foreach_False_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_clip_grad_value_foreach_True_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_conv_empty_input_cuda_complex128, test/test_nn.py::TestNNDeviceTypeCUDA::test_conv_empty_input_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_conv_empty_input_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_conv_empty_input_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_64bit_reduction_mean_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_64bit_reduction_none_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_64bit_reduction_sum_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_label_smoothing_consistent_index_target_and_probs_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_label_smoothing_errors_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_label_smoothing_weight_ignore_indices_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_label_smoothing_with_probs_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_large_tensor_reduction_mean_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_large_tensor_reduction_none_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_large_tensor_reduction_sum_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_loss_2d_out_of_bounds_class_index_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_loss_2d_out_of_bounds_class_index_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_loss_index_target_unit_weights_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_loss_one_hot_target_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_loss_prob_target_all_reductions_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_loss_prob_target_no_batch_dim_reduction_mean_weighted_False_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_loss_prob_target_no_batch_dim_reduction_mean_weighted_True_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_loss_prob_target_no_batch_dim_reduction_none_weighted_False_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_loss_prob_target_no_batch_dim_reduction_none_weighted_True_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_loss_prob_target_no_batch_dim_reduction_sum_weighted_False_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_loss_prob_target_no_batch_dim_reduction_sum_weighted_True_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_cross_entropy_loss_prob_target_unit_weights_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_ctc_loss_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_ctc_loss_cudnn_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_ctc_loss_cudnn_tensor_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_device_mask_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_elu_inplace_overlap_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_elu_inplace_with_neg_alpha_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_fold_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_glu_bfloat16_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_grid_sample_bfloat16_precision_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_grid_sample_half_precision_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_grid_sample_large_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_grid_sample_large_index_2d_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_grid_sample_large_index_2d_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_grid_sample_large_index_3d_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_grid_sample_large_index_3d_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_grid_sample_nan_inf_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_grid_sample_nan_inf_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_groupnorm_nhwc_cuda_bfloat16, test/test_nn.py::TestNNDeviceTypeCUDA::test_groupnorm_nhwc_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_groupnorm_nhwc_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_groupnorm_nhwc_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_gumbel_softmax_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_gumbel_softmax_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_gumbel_softmax_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_hardsigmoid_grad_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_hardswish_grad_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_hardswish_inplace_overlap_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_instancenorm_raises_error_for_single_spatial_element_during_training_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_instancenorm_raises_error_if_input_channels_is_not_num_features_InstanceNorm1d_no_batch_dim_False_affine_False_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_instancenorm_raises_error_if_input_channels_is_not_num_features_InstanceNorm1d_no_batch_dim_False_affine_True_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_instancenorm_raises_error_if_input_channels_is_not_num_features_InstanceNorm1d_no_batch_dim_True_affine_False_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_instancenorm_raises_error_if_input_channels_is_not_num_features_InstanceNorm1d_no_batch_dim_True_affine_True_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_instancenorm_raises_error_if_input_channels_is_not_num_features_InstanceNorm2d_no_batch_dim_False_affine_False_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_instancenorm_raises_error_if_input_channels_is_not_num_features_InstanceNorm2d_no_batch_dim_False_affine_True_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_instancenorm_raises_error_if_input_channels_is_not_num_features_InstanceNorm2d_no_batch_dim_True_affine_False_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_instancenorm_raises_error_if_input_channels_is_not_num_features_InstanceNorm2d_no_batch_dim_True_affine_True_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_instancenorm_raises_error_if_input_channels_is_not_num_features_InstanceNorm3d_no_batch_dim_False_affine_False_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_instancenorm_raises_error_if_input_channels_is_not_num_features_InstanceNorm3d_no_batch_dim_False_affine_True_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_instancenorm_raises_error_if_input_channels_is_not_num_features_InstanceNorm3d_no_batch_dim_True_affine_False_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_instancenorm_raises_error_if_input_channels_is_not_num_features_InstanceNorm3d_no_batch_dim_True_affine_True_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_instancenorm_raises_error_if_less_than_one_value_per_channel_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_invalid_reduction_strings_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_layernorm_half_precision_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_layernorm_weight_bias_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_leaky_relu_inplace_overlap_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_leaky_relu_inplace_with_neg_slope_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_leaky_relu_inplace_with_zero_slope_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_linear_empty_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_log_softmax_big_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_log_softmax_big_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_log_softmax_cpu_cuda_bfloat16, test/test_nn.py::TestNNDeviceTypeCUDA::test_log_softmax_cpu_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_logsigmoid_out_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_lstmcell_backward_only_one_output_grad_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_masked_softmax_TxT_layout_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_masked_softmax_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_masked_softmax_devices_parity_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_masked_softmax_forward_with_nans_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_masked_softmax_grad_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_masked_softmax_lowp_cuda_bfloat16, test/test_nn.py::TestNNDeviceTypeCUDA::test_masked_softmax_lowp_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_masked_softmax_mask_types_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_masked_softmax_transformer_layout_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_mish_inplace_overlap_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_module_to_empty_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_module_to_empty_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_module_to_empty_non_recursive_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nll_loss_all_ignored_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nll_loss_byte_target_matches_long_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nll_loss_empty_tensor_reduction_mean_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nll_loss_empty_tensor_reduction_none_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nll_loss_empty_tensor_reduction_sum_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nll_loss_invalid_target_dim_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nll_loss_invalid_weights_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nll_loss_large_tensor_reduction_mean_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nll_loss_large_tensor_reduction_none_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nll_loss_large_tensor_reduction_sum_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nll_loss_mismatched_batch_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nll_loss_out_of_bounds_ignore_index_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nll_loss_total_weight_is_zero_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nn_empty_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nn_scalars_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nn_scalars_reductions_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_nonlinearity_propagate_nan_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_one_hot_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_overwrite_module_params_on_conversion_cpu_device_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_pad_cuda_complex128, test/test_nn.py::TestNNDeviceTypeCUDA::test_pad_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_prelu_backward_32bit_indexing_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_replicatepad_64bit_indexing_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_rnn_fused_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_rnn_fused_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_rnn_retain_variables_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_rnn_retain_variables_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_rnn_retain_variables_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_save_lstm_compatibility_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_silu_inplace_overlap_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_skip_init_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_smooth_l1_loss_bfloat16_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_smooth_l1_loss_vs_huber_loss_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_smoothl1loss_backward_zero_beta_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_softmax_64bit_indexing_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_softmax_backward_64bit_indexing_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_softmax_bfloat16_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_softmax_cpu_cuda_bfloat16, test/test_nn.py::TestNNDeviceTypeCUDA::test_softmax_cpu_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_softmax_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_softmax_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_softmax_forward_64bit_indexing_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_softmax_results_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_softmax_results_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_softplus_inplace_overlap_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_softplus_low_threshold_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_softshrink_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_softshrink_inplace_overlap_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_softshrink_negative_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_threshold_inplace_overlap_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_to_complex_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_transformerencoderlayer_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_transformerencoderlayer_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_transformerencoderlayer_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_transformerencoderlayer_fast_path_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_transformerencoderlayer_gelu_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_transformerencoderlayer_gelu_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_triplet_margin_with_distance_loss_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_triplet_margin_with_distance_loss_default_parity_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiLinear2d_consistency_interp_size_bug_memory_format0_align_corners_False_input_size_399_output_size_437_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiLinear2d_consistency_interp_size_bug_memory_format0_align_corners_False_input_size_403_output_size_377_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiLinear2d_consistency_interp_size_bug_memory_format0_align_corners_True_input_size_399_output_size_437_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiLinear2d_consistency_interp_size_bug_memory_format0_align_corners_True_input_size_403_output_size_377_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiLinear2d_consistency_interp_size_bug_memory_format1_align_corners_False_input_size_399_output_size_437_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiLinear2d_consistency_interp_size_bug_memory_format1_align_corners_False_input_size_403_output_size_377_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiLinear2d_consistency_interp_size_bug_memory_format1_align_corners_True_input_size_399_output_size_437_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiLinear2d_consistency_interp_size_bug_memory_format1_align_corners_True_input_size_403_output_size_377_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_False_align_corners_False_mode_bicubic_memory_format0_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_False_align_corners_False_mode_bicubic_memory_format1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_False_align_corners_False_mode_bilinear_memory_format0_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_False_align_corners_False_mode_bilinear_memory_format1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_False_align_corners_True_mode_bicubic_memory_format0_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_False_align_corners_True_mode_bicubic_memory_format1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_False_align_corners_True_mode_bilinear_memory_format0_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_False_align_corners_True_mode_bilinear_memory_format1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_True_align_corners_False_mode_bicubic_memory_format0_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_True_align_corners_False_mode_bicubic_memory_format1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_True_align_corners_False_mode_bilinear_memory_format0_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_True_align_corners_False_mode_bilinear_memory_format1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_True_align_corners_True_mode_bicubic_memory_format0_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_True_align_corners_True_mode_bicubic_memory_format1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_True_align_corners_True_mode_bilinear_memory_format0_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_antialias_True_align_corners_True_mode_bilinear_memory_format1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format0_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bicubic_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_False_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_False_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_3_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_32_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_False_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_False_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_restrided_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_consistency_memory_format1_mode_bilinear_antialias_True_align_corners_True_num_channels_5_output_size_600_check_as_unsqueezed_3d_tensor_True_non_contig_sliced_batch_size_5_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_bicubic_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_bicubic_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_bicubic_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_bicubic_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_bicubic_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_bicubic_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_bicubic_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_bilinear_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_bilinear_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_bilinear_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_bilinear_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_bilinear_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_bilinear_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_bilinear_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_nearest-exact_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_nearest-exact_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_nearest-exact_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_nearest-exact_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_nearest-exact_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_nearest-exact_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_nearest-exact_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_nearest_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_nearest_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_nearest_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_nearest_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_nearest_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_nearest_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_3_mode_nearest_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_bicubic_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_bicubic_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_bicubic_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_bicubic_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_bicubic_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_bicubic_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_bicubic_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_bilinear_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_bilinear_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_bilinear_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_bilinear_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_bilinear_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_bilinear_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_bilinear_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_nearest-exact_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_nearest-exact_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_nearest-exact_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_nearest-exact_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_nearest-exact_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_nearest-exact_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_nearest-exact_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_nearest_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_nearest_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_nearest_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_nearest_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_nearest_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_nearest_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_False_num_channels_5_mode_nearest_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_bicubic_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_bicubic_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_bicubic_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_bicubic_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_bicubic_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_bicubic_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_bicubic_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_bilinear_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_bilinear_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_bilinear_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_bilinear_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_bilinear_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_bilinear_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_bilinear_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_nearest-exact_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_nearest-exact_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_nearest-exact_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_nearest-exact_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_nearest-exact_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_nearest-exact_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_nearest-exact_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_nearest_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_nearest_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_nearest_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_nearest_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_nearest_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_nearest_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_3_mode_nearest_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_bicubic_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_bicubic_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_bicubic_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_bicubic_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_bicubic_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_bicubic_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_bicubic_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_bilinear_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_bilinear_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_bilinear_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_bilinear_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_bilinear_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_bilinear_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_bilinear_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_nearest-exact_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_nearest-exact_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_nearest-exact_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_nearest-exact_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_nearest-exact_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_nearest-exact_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_nearest-exact_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_nearest_float32_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_nearest_float64_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_nearest_int16_cuda_int16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_nearest_int32_cuda_int32, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_nearest_int64_cuda_int64, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_nearest_int8_cuda_int8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBiMode2d_nonsupported_dtypes_antialias_True_num_channels_5_mode_nearest_uint8_cuda_uint8, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBicubic2d_aa_correctness_memory_format0_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBicubic2d_aa_correctness_memory_format1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBicubic2d_correctness_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBilinear2d_aa_correctness_memory_format0_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingBilinear2d_aa_correctness_memory_format1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest1d_correctness_isize_10_osize_15_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest1d_correctness_isize_20_osize_11_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest1d_launch_config_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest1d_mode_nearest-exact_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest1d_mode_nearest_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest2d_correctness_memory_format0_isize_10_osize_15_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest2d_correctness_memory_format0_isize_20_osize_11_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest2d_correctness_memory_format1_isize_10_osize_15_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest2d_correctness_memory_format1_isize_20_osize_11_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest2d_launch_config_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest2d_launch_fail_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest2d_launch_rocm_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest2d_memory_format0_mode_nearest-exact_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest2d_memory_format0_mode_nearest_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest2d_memory_format1_mode_nearest-exact_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest2d_memory_format1_mode_nearest_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest3d_correctness_memory_format0_isize_10_osize_15_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest3d_correctness_memory_format0_isize_20_osize_11_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest3d_correctness_memory_format1_isize_10_osize_15_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest3d_correctness_memory_format1_isize_20_osize_11_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest3d_launch_config_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest3d_memory_format0_mode_nearest-exact_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest3d_memory_format0_mode_nearest_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest3d_memory_format1_mode_nearest-exact_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearest3d_memory_format1_mode_nearest_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearestExact1d_correctness_isize_10_osize_15_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearestExact1d_correctness_isize_20_osize_11_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearestExact1d_rescale_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearestExact2d_correctness_memory_format0_isize_10_osize_15_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearestExact2d_correctness_memory_format0_isize_20_osize_11_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearestExact2d_correctness_memory_format1_isize_10_osize_15_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearestExact2d_correctness_memory_format1_isize_20_osize_11_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearestExact3d_correctness_memory_format0_isize_10_osize_15_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearestExact3d_correctness_memory_format0_isize_20_osize_11_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearestExact3d_correctness_memory_format1_isize_10_osize_15_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingNearestExact3d_correctness_memory_format1_isize_20_osize_11_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingTrilinear3d_align_corners_False_memory_format0_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingTrilinear3d_align_corners_False_memory_format1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingTrilinear3d_align_corners_True_memory_format0_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingTrilinear3d_align_corners_True_memory_format1_cuda, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsampling_64bit_indexing_channels_last_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_upsamplingnearest2d_backward_64bit_indexing_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_variable_sequence_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_variable_sequence_cuda_float32, test/test_nn.py::TestNNDeviceTypeCUDA::test_variable_sequence_cuda_float64, test/test_nn.py::TestNNDeviceTypeCUDA::test_warp_softmax_64bit_indexing_cuda_float16, test/test_nn.py::TestNNDeviceTypeCUDA::test_warp_softmax_64bit_indexing_cuda_float32 2024-08-06T20:45:29.7485740Z 2024-08-06T20:45:29.7486027Z Running test_overrides 1/1 ... [2024-08-06 20:45:29.487597] 2024-08-06T20:45:29.7487341Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_overrides.py', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:45:29.488121] 2024-08-06T20:45:38.0679117Z 2024-08-06T20:45:38.0680052Z test_overrides 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_overrides_1.1_365907168bbc09ff_.log 2024-08-06T20:45:38.1228942Z Running 1458 items in this shard: test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_H___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_T___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase__backward_hooks___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase__base___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase__cdata___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase__grad___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase__grad_fn___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase__post_accumulate_grad_hooks___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase__version___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_data___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_device___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_dtype___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_grad___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_grad_fn___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_imag___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_cpu___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_cuda___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_ipu___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_leaf___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_maia___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_meta___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_mkldnn___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_mps___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_mtia___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_nested___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_quantized___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_sparse___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_sparse_csr___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_vulkan___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_xla___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_is_xpu___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_itemsize___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_layout___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_mH___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_mT___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_name___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_names___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_nbytes___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_ndim___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_output_nr___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_real___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_requires_grad___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_retains_grad___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_shape___get__, test/test_overrides.py::TestTorchFunctionOverride::test_TensorBase_volatile___get__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___add__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___and__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___array__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___array_wrap__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___bool__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___complex__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___contains__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___cuda_array_interface_____get__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___deepcopy__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___div__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___dlpack__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___dlpack_device__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___eq__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___float__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___floordiv__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___format__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___ge__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___getitem__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___gt__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___iadd__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___iand__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___idiv__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___ifloordiv__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___ilshift__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___imod__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___imul__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___index__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___int__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___invert__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___ior__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___irshift__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___isub__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___ixor__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___le__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___len__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___long__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___lshift__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___lt__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___matmul__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___mod__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___mul__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___ne__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___nonzero__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___or__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___radd__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___rand__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___rdiv__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___reduce_ex__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___repr__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___reversed__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___rfloordiv__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___rlshift__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___rmatmul__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___rmod__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___rmul__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___ror__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___rpow__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___rrshift__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___rshift__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___rsub__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___rxor__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___setitem__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___setstate__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___sub__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___truediv__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor___xor__, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__autocast_to_full_precision, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__autocast_to_reduced_precision, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__coalesced_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__dimI, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__dimV, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__indices, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__is_view, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__nested_tensor_size, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__nested_tensor_storage_offsets, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__nested_tensor_strides, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__nnz, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__sparse_mask_projection, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__to_dense, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__update_names, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor__values, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_abs, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_abs_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_absolute, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_absolute_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_acos, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_acos_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_acosh, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_acosh_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_add, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_add_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_addbmm, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_addbmm_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_addcdiv, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_addcdiv_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_addcmul, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_addcmul_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_addmm, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_addmm_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_addmv, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_addmv_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_addr, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_addr_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_adjoint, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_align_as, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_align_to, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_all, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_allclose, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_amax, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_amin, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_aminmax, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_angle, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_any, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_apply_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_arccos, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_arccos_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_arccosh, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_arccosh_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_arcsin, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_arcsin_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_arcsinh, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_arcsinh_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_arctan, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_arctan2, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_arctan2_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_arctan_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_arctanh, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_arctanh_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_argmax, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_argmin, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_argsort, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_argwhere, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_as_strided, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_as_strided_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_as_strided_scatter, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_asin, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_asin_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_asinh, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_asinh_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_atan, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_atan2, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_atan2_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_atan_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_atanh, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_atanh_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_backward, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_baddbmm, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_baddbmm_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bernoulli, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bernoulli_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bfloat16, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bincount, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bitwise_and, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bitwise_and_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bitwise_left_shift, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bitwise_left_shift_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bitwise_not, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bitwise_not_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bitwise_or, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bitwise_or_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bitwise_right_shift, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bitwise_right_shift_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bitwise_xor, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bitwise_xor_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bmm, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_bool, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_broadcast_to, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_byte, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cauchy_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_ccol_indices, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cdouble, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_ceil, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_ceil_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cfloat, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_chalf, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_char, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cholesky, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cholesky_inverse, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cholesky_solve, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_chunk, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_clamp, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_clamp_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_clamp_max, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_clamp_max_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_clamp_min, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_clamp_min_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_clip, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_clip_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_clone, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_coalesce, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_col_indices, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_conj, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_conj_physical, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_conj_physical_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_contiguous, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_copy_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_copysign, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_copysign_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_corrcoef, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cos, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cos_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cosh, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cosh_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_count_nonzero, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cov, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cpu, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cross, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_crow_indices, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cuda, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cummax, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cummin, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cumprod, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cumprod_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cumsum, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_cumsum_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_data_ptr, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_deg2rad, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_deg2rad_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_dense_dim, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_dequantize, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_det, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_detach, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_detach_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_diag, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_diag_embed, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_diagflat, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_diagonal, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_diagonal_scatter, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_diff, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_digamma, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_digamma_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_dim, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_dim_order, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_dist, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_div, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_div_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_divide, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_divide_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_dot, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_double, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_dsplit, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_element_size, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_eq, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_eq_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_equal, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_erf, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_erf_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_erfc, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_erfc_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_erfinv, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_erfinv_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_exp, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_exp2, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_exp2_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_exp_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_expand, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_expand_as, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_expm1, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_expm1_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_exponential_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_fill_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_fill_diagonal_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_fix, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_fix_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_flatten, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_flip, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_fliplr, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_flipud, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_float, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_float_power, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_float_power_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_floor, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_floor_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_floor_divide, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_floor_divide_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_fmax, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_fmin, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_fmod, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_fmod_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_frac, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_frac_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_frexp, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_gather, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_gcd, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_gcd_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_ge, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_ge_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_geometric_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_geqrf, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_ger, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_get_device, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_greater, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_greater_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_greater_equal, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_greater_equal_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_gt, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_gt_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_half, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_hardshrink, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_has_names, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_heaviside, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_heaviside_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_histc, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_histogram, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_hsplit, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_hypot, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_hypot_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_i0, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_i0_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_igamma, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_igamma_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_igammac, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_igammac_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_index_add, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_index_add_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_index_copy, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_index_copy_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_index_fill, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_index_fill_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_index_put, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_index_put_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_index_reduce, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_index_reduce_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_index_select, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_indices, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_inner, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_int, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_int_repr, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_inverse, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_ipu, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_is_coalesced, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_is_complex, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_is_conj, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_is_contiguous, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_is_distributed, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_is_floating_point, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_is_inference, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_is_neg, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_is_nonzero, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_is_pinned, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_is_same_size, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_is_set_to, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_is_shared, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_is_signed, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_isclose, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_isfinite, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_isinf, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_isnan, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_isneginf, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_isposinf, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_isreal, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_istft, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_item, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_kron, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_kthvalue, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_lcm, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_lcm_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_ldexp, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_ldexp_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_le, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_le_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_lerp, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_lerp_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_less, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_less_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_less_equal, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_less_equal_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_lgamma, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_lgamma_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_log, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_log10, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_log10_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_log1p, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_log1p_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_log2, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_log2_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_log_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_log_normal_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_log_softmax, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logaddexp, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logaddexp2, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logcumsumexp, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logdet, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logical_and, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logical_and_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logical_not, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logical_not_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logical_or, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logical_or_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logical_xor, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logical_xor_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logit, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logit_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_logsumexp, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_long, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_lt, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_lt_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_lu, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_lu_solve, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_map2_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_map_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_masked_fill, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_masked_fill_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_masked_scatter, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_masked_scatter_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_masked_select, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_matmul, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_matrix_exp, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_matrix_power, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_max, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_maximum, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_mean, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_median, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_min, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_minimum, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_mm, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_mode, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_module_load, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_moveaxis, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_movedim, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_msort, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_mtia, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_mul, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_mul_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_multinomial, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_multiply, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_multiply_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_mv, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_mvlgamma, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_mvlgamma_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_nan_to_num, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_nan_to_num_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_nanmean, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_nanmedian, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_nanquantile, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_nansum, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_narrow, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_narrow_copy, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_ndimension, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_ne, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_ne_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_neg, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_neg_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_negative, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_negative_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_nelement, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_nextafter, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_nextafter_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_nonzero, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_nonzero_static, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_norm, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_normal_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_not_equal, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_not_equal_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_numel, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_numpy, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_orgqr, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_ormqr, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_outer, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_permute, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_pin_memory, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_pinverse, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_polygamma, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_polygamma_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_positive, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_pow, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_pow_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_prelu, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_prod, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_put, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_put_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_q_per_channel_axis, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_q_per_channel_scales, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_q_per_channel_zero_points, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_q_scale, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_q_zero_point, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_qr, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_qscheme, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_quantile, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_rad2deg, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_rad2deg_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_random_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_ravel, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_reciprocal, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_reciprocal_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_record_stream, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_refine_names, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_register_hook, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_register_post_accumulate_grad_hook, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_relu, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_relu_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_remainder, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_remainder_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_rename, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_rename_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_renorm, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_renorm_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_repeat, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_repeat_interleave, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_requires_grad_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_reshape, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_reshape_as, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_resize, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_resize_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_resize_as, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_resize_as_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_resize_as_sparse_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_resolve_conj, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_resolve_neg, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_retain_grad, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_roll, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_rot90, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_round, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_round_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_row_indices, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_rsqrt, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_rsqrt_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_scatter, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_scatter_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_scatter_add, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_scatter_add_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_scatter_reduce, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_scatter_reduce_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_select, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_select_scatter, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_set_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sgn, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sgn_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_share_memory_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_short, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sigmoid, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sigmoid_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sign, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sign_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_signbit, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sin, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sin_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sinc, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sinc_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sinh, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sinh_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_size, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_slice_inverse, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_slice_scatter, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_slogdet, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_smm, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_softmax, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sort, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sparse_dim, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sparse_mask, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sparse_resize_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sparse_resize_and_clear_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_split, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_split_with_sizes, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sqrt, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sqrt_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_square, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_square_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_squeeze, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_squeeze_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sspaddmm, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_std, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_stft, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_storage, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_storage_offset, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_storage_type, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sub, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sub_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_subtract, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_subtract_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sum, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_sum_to_size, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_svd, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_swapaxes, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_swapaxes_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_swapdims, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_swapdims_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_t, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_t_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_take, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_take_along_dim, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_tan, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_tan_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_tanh, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_tanh_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_tensor_split, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_tile, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_to, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_to_dense, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_to_mkldnn, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_to_sparse, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_tolist, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_topk, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_trace, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_transpose, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_transpose_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_triangular_solve, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_tril, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_tril_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_triu, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_triu_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_true_divide, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_true_divide_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_trunc, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_trunc_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_type, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_type_as, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_unbind, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_unfold, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_uniform_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_unique, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_unique_consecutive, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_unsafe_chunk, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_unsafe_split, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_unsafe_split_with_sizes, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_unsqueeze, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_unsqueeze_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_untyped_storage, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_values, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_var, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_vdot, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_view, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_view_as, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_vsplit, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_where, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_xlogy, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_xlogy_, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_xpu, test/test_overrides.py::TestTorchFunctionOverride::test_Tensor_zero_, test/test_overrides.py::TestTorchFunctionOverride::test_base, test/test_overrides.py::TestTorchFunctionOverride::test_grad, test/test_overrides.py::TestTorchFunctionOverride::test_has_torch_function_non_sequence, test/test_overrides.py::TestTorchFunctionOverride::test_mean_semantics, test/test_overrides.py::TestTorchFunctionOverride::test_mm_semantics, test/test_overrides.py::TestTorchFunctionOverride::test_pow_rpow, test/test_overrides.py::TestTorchFunctionOverride::test_precedence_semantics, test/test_overrides.py::TestTorchFunctionOverride::test_tensor_subclass_propagation, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_fft, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_fft2, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_fftn, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_fftshift, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_hfft, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_hfft2, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_hfftn, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_ifft, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_ifft2, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_ifftn, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_ifftshift, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_ihfft, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_ihfft2, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_ihfftn, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_irfft, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_irfft2, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_irfftn, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_rfft, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_rfft2, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__fft_fft_rfftn, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_cholesky, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_cholesky_ex, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_cond, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_cross, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_det, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_diagonal, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_eig, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_eigh, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_eigvals, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_eigvalsh, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_householder_product, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_inv, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_inv_ex, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_ldl_factor, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_ldl_factor_ex, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_ldl_solve, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_lstsq, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_lu, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_lu_factor, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_lu_factor_ex, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_lu_solve, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_matmul, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_matrix_exp, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_matrix_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_matrix_power, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_matrix_rank, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_multi_dot, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_pinv, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_qr, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_slogdet, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_solve, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_solve_ex, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_solve_triangular, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_svd, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_svdvals, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_tensorinv, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_tensorsolve, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_vander, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_vecdot, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__linalg_linalg_vector_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__nn_avg_pool2d, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__nn_avg_pool3d, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__nn_gelu, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__nn_linear, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__nn_log_sigmoid, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__nn_one_hot, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__nn_scaled_dot_product_attention, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__nn_softplus, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__nn_softshrink, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_airy_ai, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_bessel_j0, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_bessel_j1, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_bessel_y0, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_bessel_y1, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_chebyshev_polynomial_t, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_chebyshev_polynomial_u, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_chebyshev_polynomial_v, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_chebyshev_polynomial_w, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_digamma, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_entr, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_erf, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_erfc, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_erfcx, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_erfinv, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_exp2, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_expit, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_expm1, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_gammainc, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_gammaincc, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_gammaln, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_hermite_polynomial_h, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_hermite_polynomial_he, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_i0, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_i0e, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_i1, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_i1e, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_laguerre_polynomial_l, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_legendre_polynomial_p, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_log1p, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_log_ndtr, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_log_softmax, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_logit, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_logsumexp, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_modified_bessel_i0, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_modified_bessel_i1, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_modified_bessel_k0, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_modified_bessel_k1, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_multigammaln, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_ndtr, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_ndtri, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_polygamma, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_psi, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_round, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_scaled_modified_bessel_k0, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_scaled_modified_bessel_k1, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_shifted_chebyshev_polynomial_t, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_shifted_chebyshev_polynomial_u, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_shifted_chebyshev_polynomial_v, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_shifted_chebyshev_polynomial_w, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_sinc, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_softmax, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_spherical_bessel_j0, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_xlog1py, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_xlogy, test/test_overrides.py::TestTorchFunctionOverride::test_torch__C__special_special_zeta, test/test_overrides.py::TestTorchFunctionOverride::test_torch__assert_async, test/test_overrides.py::TestTorchFunctionOverride::test_torch__conj_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch__functional_assert_async, test/test_overrides.py::TestTorchFunctionOverride::test_torch__fw_primal_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch__indices_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch__lobpcg_lobpcg, test/test_overrides.py::TestTorchFunctionOverride::test_torch__lowrank_pca_lowrank, test/test_overrides.py::TestTorchFunctionOverride::test_torch__lowrank_svd_lowrank, test/test_overrides.py::TestTorchFunctionOverride::test_torch__make_dual_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch__native_batch_norm_legit, test/test_overrides.py::TestTorchFunctionOverride::test_torch__neg_view_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch__reshape_alias_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch__rowwise_prune, test/test_overrides.py::TestTorchFunctionOverride::test_torch__sparse_broadcast_to_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch__sym_acos, test/test_overrides.py::TestTorchFunctionOverride::test_torch__sym_asin, test/test_overrides.py::TestTorchFunctionOverride::test_torch__sym_atan, test/test_overrides.py::TestTorchFunctionOverride::test_torch__sym_cos, test/test_overrides.py::TestTorchFunctionOverride::test_torch__sym_cosh, test/test_overrides.py::TestTorchFunctionOverride::test_torch__sym_sin, test/test_overrides.py::TestTorchFunctionOverride::test_torch__sym_sinh, test/test_overrides.py::TestTorchFunctionOverride::test_torch__sym_sqrt, test/test_overrides.py::TestTorchFunctionOverride::test_torch__sym_tan, test/test_overrides.py::TestTorchFunctionOverride::test_torch__sym_tanh, test/test_overrides.py::TestTorchFunctionOverride::test_torch__values_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_abs, test/test_overrides.py::TestTorchFunctionOverride::test_torch_absolute, test/test_overrides.py::TestTorchFunctionOverride::test_torch_acos, test/test_overrides.py::TestTorchFunctionOverride::test_torch_acosh, test/test_overrides.py::TestTorchFunctionOverride::test_torch_adaptive_avg_pool1d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_adaptive_max_pool1d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_add, test/test_overrides.py::TestTorchFunctionOverride::test_torch_addbmm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_addcdiv, test/test_overrides.py::TestTorchFunctionOverride::test_torch_addcmul, test/test_overrides.py::TestTorchFunctionOverride::test_torch_addmm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_addmv, test/test_overrides.py::TestTorchFunctionOverride::test_torch_addr, test/test_overrides.py::TestTorchFunctionOverride::test_torch_adjoint, test/test_overrides.py::TestTorchFunctionOverride::test_torch_affine_grid_generator, test/test_overrides.py::TestTorchFunctionOverride::test_torch_alias_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_all, test/test_overrides.py::TestTorchFunctionOverride::test_torch_allclose, test/test_overrides.py::TestTorchFunctionOverride::test_torch_alpha_dropout, test/test_overrides.py::TestTorchFunctionOverride::test_torch_amax, test/test_overrides.py::TestTorchFunctionOverride::test_torch_amin, test/test_overrides.py::TestTorchFunctionOverride::test_torch_aminmax, test/test_overrides.py::TestTorchFunctionOverride::test_torch_angle, test/test_overrides.py::TestTorchFunctionOverride::test_torch_any, test/test_overrides.py::TestTorchFunctionOverride::test_torch_arccos, test/test_overrides.py::TestTorchFunctionOverride::test_torch_arccosh, test/test_overrides.py::TestTorchFunctionOverride::test_torch_arcsin, test/test_overrides.py::TestTorchFunctionOverride::test_torch_arcsinh, test/test_overrides.py::TestTorchFunctionOverride::test_torch_arctan, test/test_overrides.py::TestTorchFunctionOverride::test_torch_arctan2, test/test_overrides.py::TestTorchFunctionOverride::test_torch_arctanh, test/test_overrides.py::TestTorchFunctionOverride::test_torch_argmax, test/test_overrides.py::TestTorchFunctionOverride::test_torch_argmin, test/test_overrides.py::TestTorchFunctionOverride::test_torch_argsort, test/test_overrides.py::TestTorchFunctionOverride::test_torch_argwhere, test/test_overrides.py::TestTorchFunctionOverride::test_torch_as_strided_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_as_strided_scatter, test/test_overrides.py::TestTorchFunctionOverride::test_torch_asin, test/test_overrides.py::TestTorchFunctionOverride::test_torch_asinh, test/test_overrides.py::TestTorchFunctionOverride::test_torch_atan, test/test_overrides.py::TestTorchFunctionOverride::test_torch_atan2, test/test_overrides.py::TestTorchFunctionOverride::test_torch_atanh, test/test_overrides.py::TestTorchFunctionOverride::test_torch_avg_pool1d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_baddbmm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_batch_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_batch_norm_backward_elemt, test/test_overrides.py::TestTorchFunctionOverride::test_torch_batch_norm_backward_reduce, test/test_overrides.py::TestTorchFunctionOverride::test_torch_batch_norm_elemt, test/test_overrides.py::TestTorchFunctionOverride::test_torch_batch_norm_gather_stats, test/test_overrides.py::TestTorchFunctionOverride::test_torch_batch_norm_gather_stats_with_counts, test/test_overrides.py::TestTorchFunctionOverride::test_torch_batch_norm_stats, test/test_overrides.py::TestTorchFunctionOverride::test_torch_batch_norm_update_stats, test/test_overrides.py::TestTorchFunctionOverride::test_torch_bernoulli, test/test_overrides.py::TestTorchFunctionOverride::test_torch_bilinear, test/test_overrides.py::TestTorchFunctionOverride::test_torch_binary_cross_entropy_with_logits, test/test_overrides.py::TestTorchFunctionOverride::test_torch_bincount, test/test_overrides.py::TestTorchFunctionOverride::test_torch_binomial, test/test_overrides.py::TestTorchFunctionOverride::test_torch_bitwise_and, test/test_overrides.py::TestTorchFunctionOverride::test_torch_bitwise_left_shift, test/test_overrides.py::TestTorchFunctionOverride::test_torch_bitwise_not, test/test_overrides.py::TestTorchFunctionOverride::test_torch_bitwise_or, test/test_overrides.py::TestTorchFunctionOverride::test_torch_bitwise_right_shift, test/test_overrides.py::TestTorchFunctionOverride::test_torch_bitwise_xor, test/test_overrides.py::TestTorchFunctionOverride::test_torch_bmm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_broadcast_to, test/test_overrides.py::TestTorchFunctionOverride::test_torch_bucketize, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cat, test/test_overrides.py::TestTorchFunctionOverride::test_torch_ccol_indices_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_ceil, test/test_overrides.py::TestTorchFunctionOverride::test_torch_celu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_channel_shuffle, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cholesky, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cholesky_inverse, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cholesky_solve, test/test_overrides.py::TestTorchFunctionOverride::test_torch_choose_qparams_optimized, test/test_overrides.py::TestTorchFunctionOverride::test_torch_chunk, test/test_overrides.py::TestTorchFunctionOverride::test_torch_clamp, test/test_overrides.py::TestTorchFunctionOverride::test_torch_clamp_max, test/test_overrides.py::TestTorchFunctionOverride::test_torch_clamp_min, test/test_overrides.py::TestTorchFunctionOverride::test_torch_clip, test/test_overrides.py::TestTorchFunctionOverride::test_torch_clone, test/test_overrides.py::TestTorchFunctionOverride::test_torch_col_indices_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_column_stack, test/test_overrides.py::TestTorchFunctionOverride::test_torch_combinations, test/test_overrides.py::TestTorchFunctionOverride::test_torch_complex, test/test_overrides.py::TestTorchFunctionOverride::test_torch_concat, test/test_overrides.py::TestTorchFunctionOverride::test_torch_concatenate, test/test_overrides.py::TestTorchFunctionOverride::test_torch_conj, test/test_overrides.py::TestTorchFunctionOverride::test_torch_conj_physical, test/test_overrides.py::TestTorchFunctionOverride::test_torch_constant_pad_nd, test/test_overrides.py::TestTorchFunctionOverride::test_torch_conv1d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_conv2d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_conv3d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_conv_tbc, test/test_overrides.py::TestTorchFunctionOverride::test_torch_conv_transpose1d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_conv_transpose2d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_conv_transpose3d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_convolution, test/test_overrides.py::TestTorchFunctionOverride::test_torch_copysign, test/test_overrides.py::TestTorchFunctionOverride::test_torch_corrcoef, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cos, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cosh, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cosine_embedding_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cosine_similarity, test/test_overrides.py::TestTorchFunctionOverride::test_torch_count_nonzero, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cov, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cross, test/test_overrides.py::TestTorchFunctionOverride::test_torch_crow_indices_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_ctc_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cummax, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cummin, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cumprod, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cumsum, test/test_overrides.py::TestTorchFunctionOverride::test_torch_cumulative_trapezoid, test/test_overrides.py::TestTorchFunctionOverride::test_torch_deg2rad, test/test_overrides.py::TestTorchFunctionOverride::test_torch_dequantize, test/test_overrides.py::TestTorchFunctionOverride::test_torch_det, test/test_overrides.py::TestTorchFunctionOverride::test_torch_detach, test/test_overrides.py::TestTorchFunctionOverride::test_torch_detach_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_diag, test/test_overrides.py::TestTorchFunctionOverride::test_torch_diag_embed, test/test_overrides.py::TestTorchFunctionOverride::test_torch_diagflat, test/test_overrides.py::TestTorchFunctionOverride::test_torch_diagonal, test/test_overrides.py::TestTorchFunctionOverride::test_torch_diagonal_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_diagonal_scatter, test/test_overrides.py::TestTorchFunctionOverride::test_torch_diff, test/test_overrides.py::TestTorchFunctionOverride::test_torch_digamma, test/test_overrides.py::TestTorchFunctionOverride::test_torch_dist, test/test_overrides.py::TestTorchFunctionOverride::test_torch_div, test/test_overrides.py::TestTorchFunctionOverride::test_torch_divide, test/test_overrides.py::TestTorchFunctionOverride::test_torch_dot, test/test_overrides.py::TestTorchFunctionOverride::test_torch_dropout, test/test_overrides.py::TestTorchFunctionOverride::test_torch_dsmm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_dsplit, test/test_overrides.py::TestTorchFunctionOverride::test_torch_dstack, test/test_overrides.py::TestTorchFunctionOverride::test_torch_embedding, test/test_overrides.py::TestTorchFunctionOverride::test_torch_embedding_bag, test/test_overrides.py::TestTorchFunctionOverride::test_torch_empty_like, test/test_overrides.py::TestTorchFunctionOverride::test_torch_eq, test/test_overrides.py::TestTorchFunctionOverride::test_torch_equal, test/test_overrides.py::TestTorchFunctionOverride::test_torch_erf, test/test_overrides.py::TestTorchFunctionOverride::test_torch_erfc, test/test_overrides.py::TestTorchFunctionOverride::test_torch_erfinv, test/test_overrides.py::TestTorchFunctionOverride::test_torch_exp, test/test_overrides.py::TestTorchFunctionOverride::test_torch_exp2, test/test_overrides.py::TestTorchFunctionOverride::test_torch_expand_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_expm1, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fake_quantize_per_channel_affine, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fake_quantize_per_tensor_affine, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fbgemm_linear_fp16_weight, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fbgemm_linear_fp16_weight_fp32_activation, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fbgemm_linear_int8_weight, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fbgemm_linear_int8_weight_fp32_activation, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fbgemm_linear_quantize_weight, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fbgemm_pack_gemm_matrix_fp16, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fbgemm_pack_quantized_matrix, test/test_overrides.py::TestTorchFunctionOverride::test_torch_feature_alpha_dropout, test/test_overrides.py::TestTorchFunctionOverride::test_torch_feature_dropout, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fix, test/test_overrides.py::TestTorchFunctionOverride::test_torch_flatten, test/test_overrides.py::TestTorchFunctionOverride::test_torch_flip, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fliplr, test/test_overrides.py::TestTorchFunctionOverride::test_torch_flipud, test/test_overrides.py::TestTorchFunctionOverride::test_torch_float_power, test/test_overrides.py::TestTorchFunctionOverride::test_torch_floor, test/test_overrides.py::TestTorchFunctionOverride::test_torch_floor_divide, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fmax, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fmin, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fmod, test/test_overrides.py::TestTorchFunctionOverride::test_torch_frac, test/test_overrides.py::TestTorchFunctionOverride::test_torch_frexp, test/test_overrides.py::TestTorchFunctionOverride::test_torch_frobenius_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_full_like, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_atleast_1d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_atleast_2d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_atleast_3d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_block_diag, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_broadcast_tensors, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_cartesian_prod, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_cdist, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_chain_matmul, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_einsum, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_lu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_meshgrid, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_split, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_stft, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_tensordot, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_unique, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_unique_consecutive, test/test_overrides.py::TestTorchFunctionOverride::test_torch_functional_unravel_index, test/test_overrides.py::TestTorchFunctionOverride::test_torch_fused_moving_avg_obs_fake_quant, test/test_overrides.py::TestTorchFunctionOverride::test_torch_gather, test/test_overrides.py::TestTorchFunctionOverride::test_torch_gcd, test/test_overrides.py::TestTorchFunctionOverride::test_torch_ge, test/test_overrides.py::TestTorchFunctionOverride::test_torch_geqrf, test/test_overrides.py::TestTorchFunctionOverride::test_torch_ger, test/test_overrides.py::TestTorchFunctionOverride::test_torch_gradient, test/test_overrides.py::TestTorchFunctionOverride::test_torch_greater, test/test_overrides.py::TestTorchFunctionOverride::test_torch_greater_equal, test/test_overrides.py::TestTorchFunctionOverride::test_torch_grid_sampler, test/test_overrides.py::TestTorchFunctionOverride::test_torch_grid_sampler_2d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_grid_sampler_3d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_group_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_gru, test/test_overrides.py::TestTorchFunctionOverride::test_torch_gru_cell, test/test_overrides.py::TestTorchFunctionOverride::test_torch_gt, test/test_overrides.py::TestTorchFunctionOverride::test_torch_hardshrink, test/test_overrides.py::TestTorchFunctionOverride::test_torch_heaviside, test/test_overrides.py::TestTorchFunctionOverride::test_torch_hinge_embedding_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_histc, test/test_overrides.py::TestTorchFunctionOverride::test_torch_histogram, test/test_overrides.py::TestTorchFunctionOverride::test_torch_histogramdd, test/test_overrides.py::TestTorchFunctionOverride::test_torch_hsmm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_hsplit, test/test_overrides.py::TestTorchFunctionOverride::test_torch_hstack, test/test_overrides.py::TestTorchFunctionOverride::test_torch_hypot, test/test_overrides.py::TestTorchFunctionOverride::test_torch_i0, test/test_overrides.py::TestTorchFunctionOverride::test_torch_igamma, test/test_overrides.py::TestTorchFunctionOverride::test_torch_igammac, test/test_overrides.py::TestTorchFunctionOverride::test_torch_imag, test/test_overrides.py::TestTorchFunctionOverride::test_torch_index_add, test/test_overrides.py::TestTorchFunctionOverride::test_torch_index_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_index_fill, test/test_overrides.py::TestTorchFunctionOverride::test_torch_index_put, test/test_overrides.py::TestTorchFunctionOverride::test_torch_index_reduce, test/test_overrides.py::TestTorchFunctionOverride::test_torch_index_select, test/test_overrides.py::TestTorchFunctionOverride::test_torch_indices_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_inner, test/test_overrides.py::TestTorchFunctionOverride::test_torch_instance_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_int_repr, test/test_overrides.py::TestTorchFunctionOverride::test_torch_inverse, test/test_overrides.py::TestTorchFunctionOverride::test_torch_is_complex, test/test_overrides.py::TestTorchFunctionOverride::test_torch_is_conj, test/test_overrides.py::TestTorchFunctionOverride::test_torch_is_distributed, test/test_overrides.py::TestTorchFunctionOverride::test_torch_is_floating_point, test/test_overrides.py::TestTorchFunctionOverride::test_torch_is_inference, test/test_overrides.py::TestTorchFunctionOverride::test_torch_is_neg, test/test_overrides.py::TestTorchFunctionOverride::test_torch_is_nonzero, test/test_overrides.py::TestTorchFunctionOverride::test_torch_is_same_size, test/test_overrides.py::TestTorchFunctionOverride::test_torch_is_signed, test/test_overrides.py::TestTorchFunctionOverride::test_torch_isclose, test/test_overrides.py::TestTorchFunctionOverride::test_torch_isfinite, test/test_overrides.py::TestTorchFunctionOverride::test_torch_isin, test/test_overrides.py::TestTorchFunctionOverride::test_torch_isinf, test/test_overrides.py::TestTorchFunctionOverride::test_torch_isnan, test/test_overrides.py::TestTorchFunctionOverride::test_torch_isneginf, test/test_overrides.py::TestTorchFunctionOverride::test_torch_isposinf, test/test_overrides.py::TestTorchFunctionOverride::test_torch_isreal, test/test_overrides.py::TestTorchFunctionOverride::test_torch_istft, test/test_overrides.py::TestTorchFunctionOverride::test_torch_kl_div, test/test_overrides.py::TestTorchFunctionOverride::test_torch_kron, test/test_overrides.py::TestTorchFunctionOverride::test_torch_kthvalue, test/test_overrides.py::TestTorchFunctionOverride::test_torch_layer_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_lcm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_ldexp, test/test_overrides.py::TestTorchFunctionOverride::test_torch_le, test/test_overrides.py::TestTorchFunctionOverride::test_torch_lerp, test/test_overrides.py::TestTorchFunctionOverride::test_torch_less, test/test_overrides.py::TestTorchFunctionOverride::test_torch_less_equal, test/test_overrides.py::TestTorchFunctionOverride::test_torch_lgamma, test/test_overrides.py::TestTorchFunctionOverride::test_torch_log, test/test_overrides.py::TestTorchFunctionOverride::test_torch_log10, test/test_overrides.py::TestTorchFunctionOverride::test_torch_log1p, test/test_overrides.py::TestTorchFunctionOverride::test_torch_log2, test/test_overrides.py::TestTorchFunctionOverride::test_torch_log_softmax, test/test_overrides.py::TestTorchFunctionOverride::test_torch_logaddexp, test/test_overrides.py::TestTorchFunctionOverride::test_torch_logaddexp2, test/test_overrides.py::TestTorchFunctionOverride::test_torch_logcumsumexp, test/test_overrides.py::TestTorchFunctionOverride::test_torch_logdet, test/test_overrides.py::TestTorchFunctionOverride::test_torch_logical_and, test/test_overrides.py::TestTorchFunctionOverride::test_torch_logical_not, test/test_overrides.py::TestTorchFunctionOverride::test_torch_logical_or, test/test_overrides.py::TestTorchFunctionOverride::test_torch_logical_xor, test/test_overrides.py::TestTorchFunctionOverride::test_torch_logit, test/test_overrides.py::TestTorchFunctionOverride::test_torch_logsumexp, test/test_overrides.py::TestTorchFunctionOverride::test_torch_lstm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_lstm_cell, test/test_overrides.py::TestTorchFunctionOverride::test_torch_lt, test/test_overrides.py::TestTorchFunctionOverride::test_torch_lu_solve, test/test_overrides.py::TestTorchFunctionOverride::test_torch_lu_unpack, test/test_overrides.py::TestTorchFunctionOverride::test_torch_margin_ranking_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_masked_fill, test/test_overrides.py::TestTorchFunctionOverride::test_torch_masked_scatter, test/test_overrides.py::TestTorchFunctionOverride::test_torch_masked_select, test/test_overrides.py::TestTorchFunctionOverride::test_torch_matmul, test/test_overrides.py::TestTorchFunctionOverride::test_torch_matrix_exp, test/test_overrides.py::TestTorchFunctionOverride::test_torch_matrix_power, test/test_overrides.py::TestTorchFunctionOverride::test_torch_max, test/test_overrides.py::TestTorchFunctionOverride::test_torch_max_pool1d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_max_pool1d_with_indices, test/test_overrides.py::TestTorchFunctionOverride::test_torch_max_pool2d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_max_pool3d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_maximum, test/test_overrides.py::TestTorchFunctionOverride::test_torch_mean, test/test_overrides.py::TestTorchFunctionOverride::test_torch_median, test/test_overrides.py::TestTorchFunctionOverride::test_torch_min, test/test_overrides.py::TestTorchFunctionOverride::test_torch_minimum, test/test_overrides.py::TestTorchFunctionOverride::test_torch_miopen_batch_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_miopen_convolution, test/test_overrides.py::TestTorchFunctionOverride::test_torch_miopen_convolution_add_relu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_miopen_convolution_relu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_miopen_convolution_transpose, test/test_overrides.py::TestTorchFunctionOverride::test_torch_miopen_depthwise_convolution, test/test_overrides.py::TestTorchFunctionOverride::test_torch_miopen_rnn, test/test_overrides.py::TestTorchFunctionOverride::test_torch_mode, test/test_overrides.py::TestTorchFunctionOverride::test_torch_moveaxis, test/test_overrides.py::TestTorchFunctionOverride::test_torch_movedim, test/test_overrides.py::TestTorchFunctionOverride::test_torch_msort, test/test_overrides.py::TestTorchFunctionOverride::test_torch_mul, test/test_overrides.py::TestTorchFunctionOverride::test_torch_multinomial, test/test_overrides.py::TestTorchFunctionOverride::test_torch_multiply, test/test_overrides.py::TestTorchFunctionOverride::test_torch_mv, test/test_overrides.py::TestTorchFunctionOverride::test_torch_mvlgamma, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nan_to_num, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nanmean, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nanmedian, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nanquantile, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nansum, test/test_overrides.py::TestTorchFunctionOverride::test_torch_narrow, test/test_overrides.py::TestTorchFunctionOverride::test_torch_narrow_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_native_batch_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_native_channel_shuffle, test/test_overrides.py::TestTorchFunctionOverride::test_torch_native_dropout, test/test_overrides.py::TestTorchFunctionOverride::test_torch_native_group_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_native_layer_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_native_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_ne, test/test_overrides.py::TestTorchFunctionOverride::test_torch_neg, test/test_overrides.py::TestTorchFunctionOverride::test_torch_negative, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nextafter, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional__threshold, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_adaptive_avg_pool2d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_adaptive_avg_pool3d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_adaptive_max_pool1d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_adaptive_max_pool1d_with_indices, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_adaptive_max_pool2d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_adaptive_max_pool2d_with_indices, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_adaptive_max_pool3d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_adaptive_max_pool3d_with_indices, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_affine_grid, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_alpha_dropout, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_batch_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_binary_cross_entropy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_binary_cross_entropy_with_logits, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_celu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_cosine_embedding_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_cross_entropy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_ctc_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_dropout, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_dropout1d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_dropout2d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_dropout3d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_elu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_embedding, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_embedding_bag, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_feature_alpha_dropout, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_fold, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_fractional_max_pool2d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_fractional_max_pool2d_with_indices, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_fractional_max_pool3d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_fractional_max_pool3d_with_indices, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_gaussian_nll_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_glu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_grid_sample, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_group_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_gumbel_softmax, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_hardtanh, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_hinge_embedding_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_huber_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_instance_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_interpolate, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_kl_div, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_l1_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_layer_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_leaky_relu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_local_response_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_log_softmax, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_lp_pool1d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_lp_pool2d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_lp_pool3d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_margin_ranking_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_max_pool1d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_max_pool1d_with_indices, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_max_pool2d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_max_pool2d_with_indices, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_max_pool3d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_max_pool3d_with_indices, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_max_unpool1d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_max_unpool2d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_max_unpool3d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_mish, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_mse_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_multi_head_attention_forward, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_multi_margin_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_multilabel_margin_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_multilabel_soft_margin_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_nll_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_normalize, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_pad, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_poisson_nll_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_relu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_relu6, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_rms_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_rrelu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_selu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_silu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_smooth_l1_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_soft_margin_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_softmax, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_softmin, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_softsign, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_tanhshrink, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_triplet_margin_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_triplet_margin_with_distance_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_functional_unfold, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_init_constant_, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_init_kaiming_uniform_, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_init_normal_, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nn_init_uniform_, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nonzero, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nonzero_static, test/test_overrides.py::TestTorchFunctionOverride::test_torch_norm_except_dim, test/test_overrides.py::TestTorchFunctionOverride::test_torch_not_equal, test/test_overrides.py::TestTorchFunctionOverride::test_torch_nuclear_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_numel, test/test_overrides.py::TestTorchFunctionOverride::test_torch_ones_like, test/test_overrides.py::TestTorchFunctionOverride::test_torch_orgqr, test/test_overrides.py::TestTorchFunctionOverride::test_torch_ormqr, test/test_overrides.py::TestTorchFunctionOverride::test_torch_outer, test/test_overrides.py::TestTorchFunctionOverride::test_torch_pairwise_distance, test/test_overrides.py::TestTorchFunctionOverride::test_torch_pdist, test/test_overrides.py::TestTorchFunctionOverride::test_torch_permute, test/test_overrides.py::TestTorchFunctionOverride::test_torch_permute_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_pinverse, test/test_overrides.py::TestTorchFunctionOverride::test_torch_pixel_shuffle, test/test_overrides.py::TestTorchFunctionOverride::test_torch_pixel_unshuffle, test/test_overrides.py::TestTorchFunctionOverride::test_torch_poisson, test/test_overrides.py::TestTorchFunctionOverride::test_torch_poisson_nll_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_polar, test/test_overrides.py::TestTorchFunctionOverride::test_torch_polygamma, test/test_overrides.py::TestTorchFunctionOverride::test_torch_positive, test/test_overrides.py::TestTorchFunctionOverride::test_torch_pow, test/test_overrides.py::TestTorchFunctionOverride::test_torch_prelu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_prod, test/test_overrides.py::TestTorchFunctionOverride::test_torch_put, test/test_overrides.py::TestTorchFunctionOverride::test_torch_q_per_channel_axis, test/test_overrides.py::TestTorchFunctionOverride::test_torch_q_per_channel_scales, test/test_overrides.py::TestTorchFunctionOverride::test_torch_q_per_channel_zero_points, test/test_overrides.py::TestTorchFunctionOverride::test_torch_q_scale, test/test_overrides.py::TestTorchFunctionOverride::test_torch_q_zero_point, test/test_overrides.py::TestTorchFunctionOverride::test_torch_qr, test/test_overrides.py::TestTorchFunctionOverride::test_torch_quantile, test/test_overrides.py::TestTorchFunctionOverride::test_torch_quantize_per_channel, test/test_overrides.py::TestTorchFunctionOverride::test_torch_quantize_per_tensor, test/test_overrides.py::TestTorchFunctionOverride::test_torch_quantize_per_tensor_dynamic, test/test_overrides.py::TestTorchFunctionOverride::test_torch_quantized_batch_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_quantized_gru_cell, test/test_overrides.py::TestTorchFunctionOverride::test_torch_quantized_lstm_cell, test/test_overrides.py::TestTorchFunctionOverride::test_torch_quantized_max_pool1d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_quantized_max_pool2d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_quantized_max_pool3d, test/test_overrides.py::TestTorchFunctionOverride::test_torch_quantized_rnn_relu_cell, test/test_overrides.py::TestTorchFunctionOverride::test_torch_quantized_rnn_tanh_cell, test/test_overrides.py::TestTorchFunctionOverride::test_torch_rad2deg, test/test_overrides.py::TestTorchFunctionOverride::test_torch_rand_like, test/test_overrides.py::TestTorchFunctionOverride::test_torch_randint_like, test/test_overrides.py::TestTorchFunctionOverride::test_torch_randn_like, test/test_overrides.py::TestTorchFunctionOverride::test_torch_ravel, test/test_overrides.py::TestTorchFunctionOverride::test_torch_real, test/test_overrides.py::TestTorchFunctionOverride::test_torch_reciprocal, test/test_overrides.py::TestTorchFunctionOverride::test_torch_relu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_remainder, test/test_overrides.py::TestTorchFunctionOverride::test_torch_renorm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_repeat_interleave, test/test_overrides.py::TestTorchFunctionOverride::test_torch_reshape, test/test_overrides.py::TestTorchFunctionOverride::test_torch_resolve_conj, test/test_overrides.py::TestTorchFunctionOverride::test_torch_resolve_neg, test/test_overrides.py::TestTorchFunctionOverride::test_torch_rms_norm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_rnn_relu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_rnn_relu_cell, test/test_overrides.py::TestTorchFunctionOverride::test_torch_rnn_tanh, test/test_overrides.py::TestTorchFunctionOverride::test_torch_rnn_tanh_cell, test/test_overrides.py::TestTorchFunctionOverride::test_torch_roll, test/test_overrides.py::TestTorchFunctionOverride::test_torch_rot90, test/test_overrides.py::TestTorchFunctionOverride::test_torch_round, test/test_overrides.py::TestTorchFunctionOverride::test_torch_row_indices_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_row_stack, test/test_overrides.py::TestTorchFunctionOverride::test_torch_rrelu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_rsqrt, test/test_overrides.py::TestTorchFunctionOverride::test_torch_rsub, test/test_overrides.py::TestTorchFunctionOverride::test_torch_saddmm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_scatter, test/test_overrides.py::TestTorchFunctionOverride::test_torch_scatter_add, test/test_overrides.py::TestTorchFunctionOverride::test_torch_scatter_reduce, test/test_overrides.py::TestTorchFunctionOverride::test_torch_searchsorted, test/test_overrides.py::TestTorchFunctionOverride::test_torch_segment_reduce, test/test_overrides.py::TestTorchFunctionOverride::test_torch_select, test/test_overrides.py::TestTorchFunctionOverride::test_torch_select_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_select_scatter, test/test_overrides.py::TestTorchFunctionOverride::test_torch_selu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sgn, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sigmoid, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sign, test/test_overrides.py::TestTorchFunctionOverride::test_torch_signbit, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sin, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sinc, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sinh, test/test_overrides.py::TestTorchFunctionOverride::test_torch_slice_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_slice_inverse, test/test_overrides.py::TestTorchFunctionOverride::test_torch_slice_scatter, test/test_overrides.py::TestTorchFunctionOverride::test_torch_slogdet, test/test_overrides.py::TestTorchFunctionOverride::test_torch_smm, test/test_overrides.py::TestTorchFunctionOverride::test_torch_softmax, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sort, test/test_overrides.py::TestTorchFunctionOverride::test_torch_split_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_split_with_sizes, test/test_overrides.py::TestTorchFunctionOverride::test_torch_split_with_sizes_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sqrt, test/test_overrides.py::TestTorchFunctionOverride::test_torch_square, test/test_overrides.py::TestTorchFunctionOverride::test_torch_squeeze, test/test_overrides.py::TestTorchFunctionOverride::test_torch_squeeze_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_stack, test/test_overrides.py::TestTorchFunctionOverride::test_torch_std, test/test_overrides.py::TestTorchFunctionOverride::test_torch_std_mean, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sub, test/test_overrides.py::TestTorchFunctionOverride::test_torch_subtract, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sum, test/test_overrides.py::TestTorchFunctionOverride::test_torch_svd, test/test_overrides.py::TestTorchFunctionOverride::test_torch_swapaxes, test/test_overrides.py::TestTorchFunctionOverride::test_torch_swapdims, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sym_float, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sym_int, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sym_ite, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sym_max, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sym_min, test/test_overrides.py::TestTorchFunctionOverride::test_torch_sym_not, test/test_overrides.py::TestTorchFunctionOverride::test_torch_t, test/test_overrides.py::TestTorchFunctionOverride::test_torch_t_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_take, test/test_overrides.py::TestTorchFunctionOverride::test_torch_take_along_dim, test/test_overrides.py::TestTorchFunctionOverride::test_torch_tan, test/test_overrides.py::TestTorchFunctionOverride::test_torch_tanh, test/test_overrides.py::TestTorchFunctionOverride::test_torch_tensor_split, test/test_overrides.py::TestTorchFunctionOverride::test_torch_threshold, test/test_overrides.py::TestTorchFunctionOverride::test_torch_tile, test/test_overrides.py::TestTorchFunctionOverride::test_torch_topk, test/test_overrides.py::TestTorchFunctionOverride::test_torch_trace, test/test_overrides.py::TestTorchFunctionOverride::test_torch_transpose, test/test_overrides.py::TestTorchFunctionOverride::test_torch_transpose_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_trapezoid, test/test_overrides.py::TestTorchFunctionOverride::test_torch_trapz, test/test_overrides.py::TestTorchFunctionOverride::test_torch_triangular_solve, test/test_overrides.py::TestTorchFunctionOverride::test_torch_tril, test/test_overrides.py::TestTorchFunctionOverride::test_torch_triplet_margin_loss, test/test_overrides.py::TestTorchFunctionOverride::test_torch_triu, test/test_overrides.py::TestTorchFunctionOverride::test_torch_true_divide, test/test_overrides.py::TestTorchFunctionOverride::test_torch_trunc, test/test_overrides.py::TestTorchFunctionOverride::test_torch_unbind, test/test_overrides.py::TestTorchFunctionOverride::test_torch_unbind_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_unflatten, test/test_overrides.py::TestTorchFunctionOverride::test_torch_unfold_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_unsafe_chunk, test/test_overrides.py::TestTorchFunctionOverride::test_torch_unsafe_split, test/test_overrides.py::TestTorchFunctionOverride::test_torch_unsafe_split_with_sizes, test/test_overrides.py::TestTorchFunctionOverride::test_torch_unsqueeze, test/test_overrides.py::TestTorchFunctionOverride::test_torch_unsqueeze_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_values_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_var, test/test_overrides.py::TestTorchFunctionOverride::test_torch_var_mean, test/test_overrides.py::TestTorchFunctionOverride::test_torch_vdot, test/test_overrides.py::TestTorchFunctionOverride::test_torch_view_as_complex, test/test_overrides.py::TestTorchFunctionOverride::test_torch_view_as_complex_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_view_as_real, test/test_overrides.py::TestTorchFunctionOverride::test_torch_view_as_real_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_view_copy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_vsplit, test/test_overrides.py::TestTorchFunctionOverride::test_torch_vstack, test/test_overrides.py::TestTorchFunctionOverride::test_torch_where, test/test_overrides.py::TestTorchFunctionOverride::test_torch_xlogy, test/test_overrides.py::TestTorchFunctionOverride::test_torch_zeros_like, test/test_overrides.py::TestTorchFunctionOverride::test_user_implementation_raises, test/test_overrides.py::TestEinsumOverride::test_wrapper, test/test_overrides.py::TestGradCheckOverride::test_gradcheck, test/test_overrides.py::TestNamedTuple::test_max, test/test_overrides.py::TestGradNewOnesOverride::test_newones, test/test_overrides.py::TestPickle::test_pickle, test/test_overrides.py::TestBroadcastAllOverride::test_broadcast_all, test/test_overrides.py::TestWrapTorchFunction::test_wrap_torch_function, test/test_overrides.py::TestIndexing::test_getitem, test/test_overrides.py::TestIndexing::test_getitem_subclass, test/test_overrides.py::TestIndexing::test_setitem, test/test_overrides.py::TestIndexing::test_setitem_subclass, test/test_overrides.py::TestIndexing::test_setitem_val, test/test_overrides.py::TestIterator::test_iterator, test/test_overrides.py::TestRNN::test_rnn, test/test_overrides.py::TestDisabledTorchFunction::test_parameter_does_not_prevent_dispatch, test/test_overrides.py::TestResolveName::test_resolve_name, test/test_overrides.py::TestTorchFunctionWarning::test_warn_on_invalid_torch_function_standalone_class, test/test_overrides.py::TestTorchFunctionWarning::test_warn_on_invalid_torch_function_tensor_subclass, test/test_overrides.py::TestDisabledUserWarnings::test_no_implicit_user_warning_for_deprecated_functions, test/test_overrides.py::TestTorchFunctionMode::test_all_same_mode, test/test_overrides.py::TestTorchFunctionMode::test_basic, test/test_overrides.py::TestTorchFunctionMode::test_custom_device_type, test/test_overrides.py::TestTorchFunctionMode::test_disable_enable_subclass, test/test_overrides.py::TestTorchFunctionMode::test_disable_subclass_mode, test/test_overrides.py::TestTorchFunctionMode::test_disable_subclass_not_mode, test/test_overrides.py::TestTorchFunctionMode::test_distributions_bernoulli, test/test_overrides.py::TestTorchFunctionMode::test_error_using_class_method_on_mode, test/test_overrides.py::TestTorchFunctionMode::test_factory_override, test/test_overrides.py::TestTorchFunctionMode::test_get_cur_mode, test/test_overrides.py::TestTorchFunctionMode::test_get_mode_stack, test/test_overrides.py::TestTorchFunctionMode::test_getitem_call, test/test_overrides.py::TestTorchFunctionMode::test_mode_notimplemented_loop, test/test_overrides.py::TestTorchFunctionMode::test_modes_handle_first, test/test_overrides.py::TestTorchFunctionMode::test_modes_return_notimplemented, test/test_overrides.py::TestTorchFunctionMode::test_nested_modes_with_python_has_torch_function, test/test_overrides.py::TestTorchFunctionMode::test_nested_same_mode, test/test_overrides.py::TestTorchFunctionMode::test_nn_parse_to, test/test_overrides.py::TestTorchFunctionMode::test_reentrant_mode_idiom, test/test_overrides.py::TestTorchFunctionMode::test_restacking_with_ancestor, test/test_overrides.py::TestTorchFunctionMode::test_subclass_hash, test/test_overrides.py::TestTorchFunctionMode::test_with_mode, test/test_overrides.py::TestTorchFunctionMode::test_with_mode_created_separately, test/test_overrides.py::TestTorchFunctionMode::test_with_nested_modes 2024-08-06T20:45:38.1763847Z 2024-08-06T20:45:38.1764105Z Running test_cpp_api_parity 1/1 ... [2024-08-06 20:45:38.070525] 2024-08-06T20:45:38.1765445Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_cpp_api_parity.py', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:45:38.070981] 2024-08-06T20:45:59.4744930Z 2024-08-06T20:45:59.4745936Z test_cpp_api_parity 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_cpp_api_parity_1.1_d92db0df232208df_.log 2024-08-06T20:45:59.4964260Z Running 488 items in this shard: test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_BCELoss_no_batch_dim_mean, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_BCELoss_no_batch_dim_mean_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_BCELoss_no_batch_dim_none, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_BCELoss_no_batch_dim_none_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_BCELoss_no_batch_dim_sum, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_BCELoss_no_batch_dim_sum_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_BCEWithLogitsLoss_no_batch_dim_mean, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_BCEWithLogitsLoss_no_batch_dim_mean_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_BCEWithLogitsLoss_no_batch_dim_none, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_BCEWithLogitsLoss_no_batch_dim_none_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_BCEWithLogitsLoss_no_batch_dim_sum, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_BCEWithLogitsLoss_no_batch_dim_sum_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_circular_stride2_pad2, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_circular_stride2_pad2_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_dilated, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_dilated_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_groups, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_groups_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad1, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad1_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad1size1, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad1size1_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad2, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad2_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad2size1, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad2size1_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad_same, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad_same2, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad_same2_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad_same_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad_same_dilated, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad_same_dilated_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad_valid, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_pad_valid_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_reflect_stride2_pad2, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_reflect_stride2_pad2_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_replicate_stride2_pad2, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_replicate_stride2_pad2_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_stride, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_stride_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_zero_batch, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_zero_batch_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_zeros_stride2_pad2, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv1d_zeros_stride2_pad2_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_circular_stride2_pad2, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_circular_stride2_pad2_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_depthwise, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_depthwise_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_depthwise_dilated, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_depthwise_dilated_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_depthwise_padded, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_depthwise_padded_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_depthwise_strided, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_depthwise_strided_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_depthwise_with_multiplier, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_depthwise_with_multiplier_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_dilated, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_dilated_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_groups, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_groups_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_groups_thnn, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_groups_thnn_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_no_bias, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_no_bias_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_pad_same, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_pad_same_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_pad_same_dilated, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_pad_same_dilated_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_pad_valid, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_pad_valid_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_padding, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_padding_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_reflect_stride2_pad2, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_reflect_stride2_pad2_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_replicate_stride2_pad2, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_replicate_stride2_pad2_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_strided, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_strided_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_zero_batch, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_zero_batch_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_zeros_stride2_pad2, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv2d_zeros_stride2_pad2_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_1x1x1_no_bias, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_1x1x1_no_bias_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_circular_stride2_pad2, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_circular_stride2_pad2_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_dilated, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_dilated_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_dilated_strided, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_dilated_strided_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_groups, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_groups_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_no_bias, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_no_bias_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_pad_same, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_pad_same_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_pad_same_dilated, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_pad_same_dilated_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_pad_valid, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_pad_valid_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_replicate_stride2_pad2, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_replicate_stride2_pad2_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_stride, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_stride_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_stride_padding, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_stride_padding_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_zero_batch, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_zero_batch_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_zeros_stride2_pad2, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Conv3d_zeros_stride2_pad2_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose1d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose1d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose1d_dilated, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose1d_dilated_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose1d_groups, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose1d_groups_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose1d_no_bias, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose1d_no_bias_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose2d_dilated, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose2d_dilated_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose2d_groups, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose2d_groups_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose2d_no_bias, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose2d_no_bias_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose3d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose3d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose3d_dilated, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ConvTranspose3d_dilated_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_CosineEmbeddingLoss_no_batch_dim_mean, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_CosineEmbeddingLoss_no_batch_dim_mean_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_CosineEmbeddingLoss_no_batch_dim_none, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_CosineEmbeddingLoss_no_batch_dim_none_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_CosineEmbeddingLoss_no_batch_dim_sum, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_CosineEmbeddingLoss_no_batch_dim_sum_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_CrossMapLRN2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_CrossMapLRN2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Embedding, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_discontiguous, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_discontiguous_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_max, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_max_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_max_padding_idx, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_max_padding_idx_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_mean, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_mean_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_mean_padding_idx, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_mean_padding_idx_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_sparse, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_sparse_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_sum, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_sum_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_sum_padding_idx, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_EmbeddingBag_sum_padding_idx_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Embedding_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Embedding_discontiguous, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Embedding_discontiguous_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Embedding_sparse, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Embedding_sparse_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Flatten, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Flatten_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Flatten_no_batch_dim, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Flatten_no_batch_dim_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Fold, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Fold_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Fold_int_input, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Fold_int_input_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Fold_no_batch_dim_input, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Fold_no_batch_dim_input_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Fold_no_batch_dim_int_input, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Fold_no_batch_dim_int_input_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_HingeEmbeddingLoss_no_batch_dim_mean, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_HingeEmbeddingLoss_no_batch_dim_mean_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_HingeEmbeddingLoss_no_batch_dim_none, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_HingeEmbeddingLoss_no_batch_dim_none_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_HingeEmbeddingLoss_no_batch_dim_sum, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_HingeEmbeddingLoss_no_batch_dim_sum_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_LayerNorm_3d_no_affine_large_feature, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_LayerNorm_3d_no_affine_large_feature_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Linear, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Linear_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Linear_no_batch_dim, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Linear_no_batch_dim_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Linear_no_bias, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Linear_no_bias_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MarginRankingLoss_no_batch_dim_mean, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MarginRankingLoss_no_batch_dim_mean_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MarginRankingLoss_no_batch_dim_none, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MarginRankingLoss_no_batch_dim_none_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MarginRankingLoss_no_batch_dim_sum, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MarginRankingLoss_no_batch_dim_sum_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MultiLabelMarginLoss_no_batch_dim_mean, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MultiLabelMarginLoss_no_batch_dim_mean_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MultiLabelMarginLoss_no_batch_dim_none, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MultiLabelMarginLoss_no_batch_dim_none_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MultiLabelMarginLoss_no_batch_dim_sum, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MultiLabelMarginLoss_no_batch_dim_sum_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MultiLabelSoftMarginLoss_no_batch_dim_mean, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MultiLabelSoftMarginLoss_no_batch_dim_mean_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MultiLabelSoftMarginLoss_no_batch_dim_none, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MultiLabelSoftMarginLoss_no_batch_dim_none_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MultiLabelSoftMarginLoss_no_batch_dim_sum, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_MultiLabelSoftMarginLoss_no_batch_dim_sum_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_NLLLoss_no_batch_dim_mean, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_NLLLoss_no_batch_dim_mean_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_NLLLoss_no_batch_dim_none, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_NLLLoss_no_batch_dim_none_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_NLLLoss_no_batch_dim_sum, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_NLLLoss_no_batch_dim_sum_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_PairwiseDistance, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_PairwiseDistance_broadcast_lhs, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_PairwiseDistance_broadcast_lhs_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_PairwiseDistance_broadcast_rhs, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_PairwiseDistance_broadcast_rhs_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_PairwiseDistance_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_PairwiseDistance_no_batch_dim, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_PairwiseDistance_no_batch_dim_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_PairwiseDistance_with_non_default_args, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_PairwiseDistance_with_non_default_args_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_PixelShuffle, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_PixelShuffle_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_PixelUnshuffle, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_PixelUnshuffle_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_RReLU, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_RReLU_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_RReLU_with_up_down, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_RReLU_with_up_down_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_RReLU_with_up_down_scalar, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_RReLU_with_up_down_scalar_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ReplicationPad3d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ReplicationPad3d_complex, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ReplicationPad3d_complex_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ReplicationPad3d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ReplicationPad3d_no_batch_dim, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_ReplicationPad3d_no_batch_dim_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_SampleModule_has_parity, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_SampleModule_has_parity_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_SampleModule_no_parity, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_SampleModule_no_parity_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_SoftMarginLoss_no_batch_dim_mean, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_SoftMarginLoss_no_batch_dim_mean_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_SoftMarginLoss_no_batch_dim_none, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_SoftMarginLoss_no_batch_dim_none_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_SoftMarginLoss_no_batch_dim_sum, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_SoftMarginLoss_no_batch_dim_sum_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_TransformerDecoderLayer_gelu_activation, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_TransformerDecoderLayer_gelu_activation_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_TransformerDecoderLayer_relu_activation, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_TransformerDecoderLayer_relu_activation_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_TransformerEncoderLayer_gelu_activation, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_TransformerEncoderLayer_gelu_activation_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_TransformerEncoderLayer_relu_activation, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_TransformerEncoderLayer_relu_activation_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Transformer_multilayer_coder, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Transformer_multilayer_coder_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_TripletMarginLoss_no_batch_dim_mean, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_TripletMarginLoss_no_batch_dim_mean_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_TripletMarginLoss_no_batch_dim_none, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_TripletMarginLoss_no_batch_dim_none_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_TripletMarginLoss_no_batch_dim_sum, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_TripletMarginLoss_no_batch_dim_sum_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Unflatten_no_batch_dim, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Unflatten_no_batch_dim_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Unfold, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Unfold_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Unfold_int_input, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_Unfold_int_input_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_BCELoss_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_BCELoss_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_BCELoss_no_reduce_scalar, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_BCELoss_no_reduce_scalar_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_BCELoss_weights_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_BCELoss_weights_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_BCELoss_weights_no_reduce_scalar, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_BCELoss_weights_no_reduce_scalar_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_BCEWithLogitsLoss_legacy_enum, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_BCEWithLogitsLoss_legacy_enum_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_BCEWithLogitsLoss_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_BCEWithLogitsLoss_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_BCEWithLogitsLoss_no_reduce_scalar, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_BCEWithLogitsLoss_no_reduce_scalar_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_HingeEmbeddingLoss_margin_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_HingeEmbeddingLoss_margin_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_HingeEmbeddingLoss_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_HingeEmbeddingLoss_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_HuberLoss_delta, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_HuberLoss_delta_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_KLDivLoss_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_KLDivLoss_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_KLDivLoss_no_reduce_log_target, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_KLDivLoss_no_reduce_log_target_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_KLDivLoss_no_reduce_scalar, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_KLDivLoss_no_reduce_scalar_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_KLDivLoss_no_reduce_scalar_log_target, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_KLDivLoss_no_reduce_scalar_log_target_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_KLDivLoss_with_log_target_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_KLDivLoss_with_log_target_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_KLDivLoss_with_target_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_KLDivLoss_with_target_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_L1Loss_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_L1Loss_no_reduce_complex, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_L1Loss_no_reduce_complex_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_L1Loss_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_L1Loss_no_reduce_scalar, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_L1Loss_no_reduce_scalar_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MSELoss_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MSELoss_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MSELoss_no_reduce_scalar, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MSELoss_no_reduce_scalar_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiLabelMarginLoss_0d_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiLabelMarginLoss_0d_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiLabelMarginLoss_1d_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiLabelMarginLoss_1d_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiLabelMarginLoss_index_neg, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiLabelMarginLoss_index_neg_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiLabelMarginLoss_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiLabelMarginLoss_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiLabelSoftMarginLoss_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiLabelSoftMarginLoss_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiLabelSoftMarginLoss_weights_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiLabelSoftMarginLoss_weights_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiMarginLoss_1d_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiMarginLoss_1d_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiMarginLoss_margin_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiMarginLoss_margin_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiMarginLoss_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiMarginLoss_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiMarginLoss_p_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiMarginLoss_p_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiMarginLoss_weights_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_MultiMarginLoss_weights_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss2d_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss2d_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss2d_no_reduce_ignore_index, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss2d_no_reduce_ignore_index_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss2d_no_reduce_weights, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss2d_no_reduce_weights_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLossNd_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLossNd_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLossNd_no_reduce_ignore_index, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLossNd_no_reduce_ignore_index_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLossNd_no_reduce_weights, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLossNd_no_reduce_weights_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss_no_reduce_ignore_index, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss_no_reduce_ignore_index_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss_no_reduce_weights, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss_no_reduce_weights_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss_no_reduce_weights_ignore_index, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss_no_reduce_weights_ignore_index_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss_no_reduce_weights_ignore_index_neg, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_NLLLoss_no_reduce_weights_ignore_index_neg_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_PoissonNLLLoss_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_PoissonNLLLoss_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_SmoothL1Loss_beta, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_SmoothL1Loss_beta_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_SmoothL1Loss_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_SmoothL1Loss_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_SmoothL1Loss_no_reduce_scalar, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_SmoothL1Loss_no_reduce_scalar_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_SmoothL1Loss_zero_beta, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_SmoothL1Loss_zero_beta_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_SoftMarginLoss_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_SoftMarginLoss_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_2d_zero_dim, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_2d_zero_dim_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_scale_2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_scale_2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_scale_tuple_shared_2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_scale_tuple_shared_2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_scale_tuple_skewed_2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_scale_tuple_skewed_2d_align_corners, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_scale_tuple_skewed_2d_align_corners_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_scale_tuple_skewed_2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_tuple_2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_tuple_2d_align_corners, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_tuple_2d_align_corners_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bicubic_tuple_2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_2d_zero_dim, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_2d_zero_dim_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_scale_2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_scale_2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_scale_tuple_shared_2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_scale_tuple_shared_2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_scale_tuple_skewed_2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_scale_tuple_skewed_2d_align_corners, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_scale_tuple_skewed_2d_align_corners_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_scale_tuple_skewed_2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_tuple_2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_tuple_2d_align_corners, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_tuple_2d_align_corners_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_bilinear_tuple_2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_linear_1d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_linear_1d_align_corners, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_linear_1d_align_corners_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_linear_1d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_linear_1d_zero_dim, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_linear_1d_zero_dim_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_linear_scale_1d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_linear_scale_1d_align_corners, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_linear_scale_1d_align_corners_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_linear_scale_1d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_linear_tuple_1d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_linear_tuple_1d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_1d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_1d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_1d_zero_dim, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_1d_zero_dim_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_2d_launch_configs, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_2d_launch_configs_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_2d_zero_dim, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_2d_zero_dim_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_3d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_3d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_3d_zero_dim, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_3d_zero_dim_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_scale_1d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_scale_1d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_scale_2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_scale_2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_scale_3d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_scale_3d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_tuple_1d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_tuple_1d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_tuple_2d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_tuple_2d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_tuple_3d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_nearest_tuple_3d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_trilinear_3d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_trilinear_3d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_trilinear_3d_zero_dim, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_trilinear_3d_zero_dim_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_trilinear_scale_3d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_trilinear_scale_3d_align_corners, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_trilinear_scale_3d_align_corners_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_trilinear_scale_3d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_trilinear_tuple_3d, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_trilinear_tuple_3d_align_corners, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_trilinear_tuple_3d_align_corners_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_interpolate_trilinear_tuple_3d_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_log_softmax_dim0, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_log_softmax_dim0_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_log_softmax_dim3, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_log_softmax_dim3_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_log_softmax_lastdim, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_log_softmax_lastdim_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_log_softmax_scalar, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_log_softmax_scalar_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_log_softmax_spatial, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_log_softmax_spatial_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_log_softmax_spatial_special, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_log_softmax_spatial_special_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_multimarginloss_1d_input_0d_target_no_reduce, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_multimarginloss_1d_input_0d_target_no_reduce_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_sample_functional_has_parity, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_sample_functional_has_parity_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_sample_functional_no_parity, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_sample_functional_no_parity_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_functional_dim0, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_functional_dim0_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_functional_dim3, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_functional_dim3_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_functional_scalar, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_functional_scalar_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_lastdim, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_lastdim_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_lastdim_dtype, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_lastdim_dtype_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_spatial, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_spatial_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_spatial_dtype, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_spatial_dtype_cuda, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_spatial_special, test/test_cpp_api_parity.py::TestCppApiParity::test_torch_nn_functional_softmax_spatial_special_cuda 2024-08-06T20:45:59.5177207Z 2024-08-06T20:45:59.5177430Z Running test_torch 1/1 ... [2024-08-06 20:45:59.475780] 2024-08-06T20:45:59.5178692Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_torch.py', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:45:59.476256] 2024-08-06T20:51:45.8208647Z 2024-08-06T20:51:45.8209887Z test_torch 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_torch_1.1_38c76b710a46fc3c_.log 2024-08-06T20:51:45.8585044Z Running 1042 items in this shard: test/test_torch.py::TestBasicVitalSigns::test_basic_vitals, test/test_torch.py::TestBasicVitalSigns::test_basic_vitals_read_write, test/test_torch.py::TestBasicVitalSigns::test_dataloader_vitals, test/test_torch.py::TestTorch::test_RNGState, test/test_torch.py::TestTorch::test_RNGStateAliasing, test/test_torch.py::TestTorch::test_RNG_after_pickle, test/test_torch.py::TestTorch::test_Size, test/test_torch.py::TestTorch::test_Size_iter, test/test_torch.py::TestTorch::test_Size_scalar, test/test_torch.py::TestTorch::test_add_meta_scalar, test/test_torch.py::TestTorch::test_allow_tensor_metadata_change, test/test_torch.py::TestTorch::test_apply, test/test_torch.py::TestTorch::test_as_subclass, test/test_torch.py::TestTorch::test_assert_async, test/test_torch.py::TestTorch::test_backward_hooks_traverse, test/test_torch.py::TestTorch::test_batch_norm_cpu_inference, test/test_torch.py::TestTorch::test_bf16_supported_on_cpu, test/test_torch.py::TestTorch::test_bmm_multithreaded, test/test_torch.py::TestTorch::test_boxMullerState, test/test_torch.py::TestTorch::test_cat_neg_dim, test/test_torch.py::TestTorch::test_check, test/test_torch.py::TestTorch::test_chunk_neg_dim, test/test_torch.py::TestTorch::test_conj_neg_tolist, test/test_torch.py::TestTorch::test_contains, test/test_torch.py::TestTorch::test_copy_broadcast, test/test_torch.py::TestTorch::test_copy_dtypes, test/test_torch.py::TestTorch::test_copy_float16, test/test_torch.py::TestTorch::test_copy_many_to_one, test/test_torch.py::TestTorch::test_copy_transpose, test/test_torch.py::TestTorch::test_cuda_not_built, test/test_torch.py::TestTorch::test_cummax_neg_dim, test/test_torch.py::TestTorch::test_cummin_neg_dim, test/test_torch.py::TestTorch::test_cumprod_neg_dim, test/test_torch.py::TestTorch::test_cumsum_neg_dim, test/test_torch.py::TestTorch::test_cxx_flags, test/test_torch.py::TestTorch::test_data_ptr_of_empty_tensor_with_storage, test/test_torch.py::TestTorch::test_data_ptr_of_empty_view_with_storage, test/test_torch.py::TestTorch::test_deepcopy_gradient, test/test_torch.py::TestTorch::test_deepcopy_parameter, test/test_torch.py::TestTorch::test_deterministic_fill_uninitialized_memory, test/test_torch.py::TestTorch::test_deterministic_flag, test/test_torch.py::TestTorch::test_device, test/test_torch.py::TestTorch::test_dim_order, test/test_torch.py::TestTorch::test_dir, test/test_torch.py::TestTorch::test_doc, test/test_torch.py::TestTorch::test_doc_template, test/test_torch.py::TestTorch::test_dot_data_use, test/test_torch.py::TestTorch::test_dtype_is_signed, test/test_torch.py::TestTorch::test_element_size, test/test_torch.py::TestTorch::test_empty_meta, test/test_torch.py::TestTorch::test_empty_storage_view, test/test_torch.py::TestTorch::test_equal, test/test_torch.py::TestTorch::test_error_msg_type_translation, test/test_torch.py::TestTorch::test_fill_diagonal, test/test_torch.py::TestTorch::test_format_scalar_meta, test/test_torch.py::TestTorch::test_from_buffer, test/test_torch.py::TestTorch::test_from_file, test/test_torch.py::TestTorch::test_gather_neg_dim, test/test_torch.py::TestTorch::test_generator_cpu, test/test_torch.py::TestTorch::test_get_cpu_capability, test/test_torch.py::TestTorch::test_has_internal_overlap, test/test_torch.py::TestTorch::test_has_storage, test/test_torch.py::TestTorch::test_index_add, test/test_torch.py::TestTorch::test_index_add_all_dtypes, test/test_torch.py::TestTorch::test_index_add_cornercase, test/test_torch.py::TestTorch::test_index_add_correctness, test/test_torch.py::TestTorch::test_index_add_neg_dim, test/test_torch.py::TestTorch::test_index_copy_neg_dim, test/test_torch.py::TestTorch::test_index_fill_neg_dim, test/test_torch.py::TestTorch::test_index_select_neg_dim, test/test_torch.py::TestTorch::test_invalid_arg_error_handling, test/test_torch.py::TestTorch::test_invalid_generator_raises, test/test_torch.py::TestTorch::test_is_nonzero, test/test_torch.py::TestTorch::test_is_same_size, test/test_torch.py::TestTorch::test_iter, test/test_torch.py::TestTorch::test_kthvalue_neg_dim, test/test_torch.py::TestTorch::test_linspace_logspace, test/test_torch.py::TestTorch::test_logcumsumexp_neg_dim, test/test_torch.py::TestTorch::test_manual_seed, test/test_torch.py::TestTorch::test_map, test/test_torch.py::TestTorch::test_map2, test/test_torch.py::TestTorch::test_max_neg_dim, test/test_torch.py::TestTorch::test_mean_neg_dim, test/test_torch.py::TestTorch::test_median_neg_dim, test/test_torch.py::TestTorch::test_memory_format, test/test_torch.py::TestTorch::test_memory_format_contiguous_returns_same_tensor_if_already_satisfies, test/test_torch.py::TestTorch::test_memory_format_empty, test/test_torch.py::TestTorch::test_min_neg_dim, test/test_torch.py::TestTorch::test_mode_neg_dim, test/test_torch.py::TestTorch::test_multinomial_invalid_probs, test/test_torch.py::TestTorch::test_nanmedian_neg_dim, test/test_torch.py::TestTorch::test_narrow_neg_dim, test/test_torch.py::TestTorch::test_nbytes, test/test_torch.py::TestTorch::test_ndim, test/test_torch.py::TestTorch::test_new, test/test_torch.py::TestTorch::test_newaxis_numpy_comparison, test/test_torch.py::TestTorch::test_newindex, test/test_torch.py::TestTorch::test_no_cuda_monkeypatch, test/test_torch.py::TestTorch::test_norm_neg_dim, test/test_torch.py::TestTorch::test_normal_shape, test/test_torch.py::TestTorch::test_numel, test/test_torch.py::TestTorch::test_parallel_info, test/test_torch.py::TestTorch::test_parsing_double, test/test_torch.py::TestTorch::test_parsing_int64, test/test_torch.py::TestTorch::test_parsing_intlist, test/test_torch.py::TestTorch::test_permute, test/test_torch.py::TestTorch::test_pickle, test/test_torch.py::TestTorch::test_pickle_dtype, test/test_torch.py::TestTorch::test_pickle_function, test/test_torch.py::TestTorch::test_pickle_generator, test/test_torch.py::TestTorch::test_pickle_parameter, test/test_torch.py::TestTorch::test_pickle_parameter_no_requires_grad, test/test_torch.py::TestTorch::test_pickle_size, test/test_torch.py::TestTorch::test_pin_memory, test/test_torch.py::TestTorch::test_print, test/test_torch.py::TestTorch::test_prod_neg_dim, test/test_torch.py::TestTorch::test_pyobj_preserved, test/test_torch.py::TestTorch::test_qengine, test/test_torch.py::TestTorch::test_renorm_neg_dim, test/test_torch.py::TestTorch::test_resizable, test/test_torch.py::TestTorch::test_reversed, test/test_torch.py::TestTorch::test_scatter_neg_dim, test/test_torch.py::TestTorch::test_select_neg_dim, test/test_torch.py::TestTorch::test_set_flush_denormal, test/test_torch.py::TestTorch::test_setting_real_imag_to_a_number, test/test_torch.py::TestTorch::test_show_config, test/test_torch.py::TestTorch::test_size_neg_dim, test/test_torch.py::TestTorch::test_size_stride, test/test_torch.py::TestTorch::test_sizeof, test/test_torch.py::TestTorch::test_slice, test/test_torch.py::TestTorch::test_slow_test, test/test_torch.py::TestTorch::test_sobolengine_bounds, test/test_torch.py::TestTorch::test_sobolengine_bounds_scrambled, test/test_torch.py::TestTorch::test_sobolengine_continuing, test/test_torch.py::TestTorch::test_sobolengine_continuing_scrambled, test/test_torch.py::TestTorch::test_sobolengine_default_dtype, test/test_torch.py::TestTorch::test_sobolengine_distribution, test/test_torch.py::TestTorch::test_sobolengine_distribution_scrambled, test/test_torch.py::TestTorch::test_sobolengine_draw, test/test_torch.py::TestTorch::test_sobolengine_draw_base2, test/test_torch.py::TestTorch::test_sobolengine_draw_base2_scrambled, test/test_torch.py::TestTorch::test_sobolengine_draw_scrambled, test/test_torch.py::TestTorch::test_sobolengine_fast_forward, test/test_torch.py::TestTorch::test_sobolengine_fast_forward_scrambled, test/test_torch.py::TestTorch::test_sobolengine_first_point, test/test_torch.py::TestTorch::test_sobolengine_high_dim, test/test_torch.py::TestTorch::test_sobolengine_raise, test/test_torch.py::TestTorch::test_sobolengine_reset, test/test_torch.py::TestTorch::test_sobolengine_reset_scrambled, test/test_torch.py::TestTorch::test_sort_neg_dim, test/test_torch.py::TestTorch::test_split_neg_dim, test/test_torch.py::TestTorch::test_split_with_sizes_copy_out, test/test_torch.py::TestTorch::test_squeeze_neg_dim, test/test_torch.py::TestTorch::test_std_neg_dim, test/test_torch.py::TestTorch::test_storage_base_init, test/test_torch.py::TestTorch::test_storage_base_new, test/test_torch.py::TestTorch::test_storage_byteswap, test/test_torch.py::TestTorch::test_storage_casts, test/test_torch.py::TestTorch::test_storage_cycle_via_dict, test/test_torch.py::TestTorch::test_storage_cycle_via_slots, test/test_torch.py::TestTorch::test_storage_dead_weak_ref, test/test_torch.py::TestTorch::test_storage_dealloc, test/test_torch.py::TestTorch::test_storage_dealloc_resurrected, test/test_torch.py::TestTorch::test_storage_dealloc_subclass_resurrected, test/test_torch.py::TestTorch::test_storage_dealloc_subclass_zombie, test/test_torch.py::TestTorch::test_storage_dict_dealloc, test/test_torch.py::TestTorch::test_storage_error, test/test_torch.py::TestTorch::test_storage_error_no_attribute, test/test_torch.py::TestTorch::test_storage_finalizer_dealloc, test/test_torch.py::TestTorch::test_storage_fix_weakref_no_leak, test/test_torch.py::TestTorch::test_storage_from_tensor_dealloc, test/test_torch.py::TestTorch::test_storage_from_tensor_dealloc_resurrected, test/test_torch.py::TestTorch::test_storage_from_tensor_dealloc_zombie, test/test_torch.py::TestTorch::test_storage_preserve_nonhermetic_in_hermetic_context, test/test_torch.py::TestTorch::test_storage_resurrected_weak_ref, test/test_torch.py::TestTorch::test_storage_slot_dealloc, test/test_torch.py::TestTorch::test_storage_weakref_dealloc, test/test_torch.py::TestTorch::test_structseq_repr, test/test_torch.py::TestTorch::test_subclass_preserved, test/test_torch.py::TestTorch::test_subclass_tensors, test/test_torch.py::TestTorch::test_sum_neg_dim, test/test_torch.py::TestTorch::test_swap_basic, test/test_torch.py::TestTorch::test_swap_fail_slots, test/test_torch.py::TestTorch::test_t_not_2d_error, test/test_torch.py::TestTorch::test_tensor_base_init, test/test_torch.py::TestTorch::test_tensor_base_new, test/test_torch.py::TestTorch::test_tensor_ctor_scalar, test/test_torch.py::TestTorch::test_tensor_cycle_via_dict, test/test_torch.py::TestTorch::test_tensor_cycle_via_slots, test/test_torch.py::TestTorch::test_tensor_dead_weak_ref, test/test_torch.py::TestTorch::test_tensor_dict_dealloc, test/test_torch.py::TestTorch::test_tensor_finalizer_dealloc, test/test_torch.py::TestTorch::test_tensor_fix_weakref_no_leak, test/test_torch.py::TestTorch::test_tensor_resurrected_weak_ref, test/test_torch.py::TestTorch::test_tensor_set, test/test_torch.py::TestTorch::test_tensor_set_errors, test/test_torch.py::TestTorch::test_tensor_slot_dealloc, test/test_torch.py::TestTorch::test_tensor_weakref_dealloc, test/test_torch.py::TestTorch::test_tensor_where_scalar, test/test_torch.py::TestTorch::test_tensoriterator_output_setup, test/test_torch.py::TestTorch::test_terminate_handler_on_crash, test/test_torch.py::TestTorch::test_to, test/test_torch.py::TestTorch::test_to_with_tensor, test/test_torch.py::TestTorch::test_topk_neg_dim, test/test_torch.py::TestTorch::test_torch_from_file, test/test_torch.py::TestTorch::test_transpose_neg_dim, test/test_torch.py::TestTorch::test_type, test/test_torch.py::TestTorch::test_type_alias, test/test_torch.py::TestTorch::test_type_conversion_via_dtype_name, test/test_torch.py::TestTorch::test_typed_storage_deprecation_warning, test/test_torch.py::TestTorch::test_typed_storage_internal_no_warning, test/test_torch.py::TestTorch::test_unbind_neg_dim, test/test_torch.py::TestTorch::test_unflatten, test/test_torch.py::TestTorch::test_unfold_neg_dim, test/test_torch.py::TestTorch::test_unsqueeze_neg_dim, test/test_torch.py::TestTorch::test_upsample_nearest1d_meta, test/test_torch.py::TestTorch::test_upsample_nearest2d_meta, test/test_torch.py::TestTorch::test_var_neg_dim, test/test_torch.py::TestTorch::test_warn_types, test/test_torch.py::TestTorch::test_wildcard_import, test/test_torch.py::TestVitalSignsCudaCUDA::test_cuda_vitals_gpu_only_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcdiv_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcdiv_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcdiv_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcdiv_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcdiv_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcdiv_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcdiv_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcdiv_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcdiv_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcdiv_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcmul_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcmul_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcmul_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcmul_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcmul_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcmul_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcmul_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcmul_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcmul_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_addcmul_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_assertRaisesRegex_ignore_msg_non_native_device_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_edge_cases_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_edge_cases_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_edge_cases_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_mem_overlap_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_p_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_p_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_p_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_self_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_self_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_self_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_self_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_self_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_self_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_self_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_self_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bernoulli_self_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bfloat16_neg_abs_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bool_tensor_value_change_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_add_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_addcdiv_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_addcmul_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_atan2_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_copy_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_dist_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_div_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_eq_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_fmod_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_ge_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_gt_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_le_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_lerp_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_lt_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_map2_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_map_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_masked_fill_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_masked_scatter_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_masked_select_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_max_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_min_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_mul_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_ne_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_pow_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_remainder_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_broadcast_fn_sub_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bytes_to_scalar_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bytes_to_scalar_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bytes_to_scalar_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bytes_to_scalar_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bytes_to_scalar_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bytes_to_scalar_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bytes_to_scalar_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bytes_to_scalar_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bytes_to_scalar_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bytes_to_scalar_cuda_uint16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bytes_to_scalar_cuda_uint32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bytes_to_scalar_cuda_uint64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_bytes_to_scalar_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cauchy_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cauchy_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cauchy_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cauchy_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cauchy_kstest_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cauchy_no_inf_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cauchy_no_inf_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cdist_cuda_backward_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cdist_empty_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cdist_euclidean_large_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cdist_grad_p_lt_1_no_nan_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cdist_large_batch_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cdist_large_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cdist_non_contiguous_batch_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cdist_non_contiguous_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cdist_norm_batch_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cdist_norm_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cdist_same_inputs_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_check_tensor_all_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_check_tensor_internal_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_clone_all_dtypes_and_devices_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_clone_not_memory_dense_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_clone_zero_stride_dim_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_complex_half_experimental_warning_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_constants_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_conv_transposed_backward_agnostic_to_memory_format_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_conv_transposed_large_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy__cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy__cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy__cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy__cuda_complex32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy__cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy__cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy__cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy__cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy__cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy__cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy__cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy__cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy__cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy_all_dtypes_and_devices_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy_math_view_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy_mem_overlap_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy_transpose_math_view_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy_transpose_math_view_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_copy_transpose_math_view_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_corrcoef_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_corrcoef_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_corrcoef_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cov_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cpp_warnings_have_python_context_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cublas_config_nondeterministic_alert_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cummax_cummin_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cummax_discontiguous_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cummin_discontiguous_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cumprod_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_cumsum_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deepcopy_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deepcopy_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deepcopy_scalar_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deepcopy_scalar_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_uint16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_uint32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_uint64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_empty_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_interpolate_bilinear_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_replication_pad2d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_uint16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_uint32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_uint64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_deterministic_resize_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_device_guard_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_noncontig_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_noncontig_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_noncontig_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_noncontig_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_noncontig_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_noncontig_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_noncontig_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_noncontig_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_noncontig_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_noncontig_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_diff_noncontig_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_dim_function_empty_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_discontiguous_out_cumsum_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_dist_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_dtypetensor_warnings_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_errors_index_copy_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_expected_failure_xla_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_exponential_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_exponential_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_exponential_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_exponential_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_exponential_kstest_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_exponential_kstest_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_exponential_kstest_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_exponential_kstest_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_exponential_no_zero_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_exponential_no_zero_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_gather_backward_deterministic_path_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_gather_backward_one_dim_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_geometric_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_geometric_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_geometric_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_geometric_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_geometric_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_geometric_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_geometric_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_geometric_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_geometric_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_geometric_kstest_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scale_will_not_overflow_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaler_deprecated_warning_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaler_pass_itself_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_accumulation_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_autocast_foreach0_fused0_AdamW_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_autocast_foreach0_fused0_Adam_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_autocast_foreach0_fused0_SGD_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_autocast_foreach2_fused_True_AdamW_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_autocast_foreach2_fused_True_Adam_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_autocast_foreach2_fused_True_SGD_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_autocast_foreach_True_fused1_AdamW_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_autocast_foreach_True_fused1_Adam_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_autocast_foreach_True_fused1_SGD_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_clipping_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_clipping_separate_unscale_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_multiple_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_penalty_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_state_dict_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_unscale_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_unscale_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_unscale_sparse_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_grad_scaling_update_scale_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_gradient_all_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_gradient_all_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_gradient_all_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_gradient_extreme_cases_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_gradient_extreme_cases_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_gradient_extreme_cases_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_gradient_spacing_list_length_error_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_gradient_spacing_list_length_error_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_gradient_spacing_list_length_error_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_gradient_type_promotion_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_hook_remove_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_add_deterministic_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_add_mem_overlap_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_deterministic_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_mem_overlap_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_scalars_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_scalars_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_scalars_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_scalars_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_scalars_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_scalars_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_scalars_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_scalars_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_scalars_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_scalars_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_scalars_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_copy_scalars_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_fill_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_fill_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_fill_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_fill_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_fill_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_fill_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_fill_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_fill_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_fill_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_fill_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_fill_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_fill_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_fill_mem_overlap_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_put_mem_overlap_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_put_non_accumulate_deterministic_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amax_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amax_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amax_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amax_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amax_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amax_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amax_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amax_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amax_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amin_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amin_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amin_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amin_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amin_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amin_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amin_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amin_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_amin_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_mean_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_mean_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_mean_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_mean_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_mean_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_mean_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_mean_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_mean_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_mean_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_prod_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_prod_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_prod_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_prod_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_prod_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_prod_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_prod_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_prod_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_reduce_reduce_prod_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_select_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_select_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_select_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_select_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_select_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_select_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_select_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_select_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_select_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_select_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_select_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_index_select_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_invalid_shapes_grid_sampler_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_is_set_to_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_is_signed_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_complex32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_uint16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_uint32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_uint64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_item_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_large_cumprod_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_large_cumsum_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_binary_op_no_materialize_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_binary_op_no_materialize_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_binary_op_no_materialize_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_binary_op_no_materialize_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_binary_op_no_materialize_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_binary_op_no_materialize_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_binary_op_no_materialize_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_binary_op_no_materialize_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_binary_op_no_materialize_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_binary_op_no_materialize_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_binary_op_no_materialize_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_binary_op_no_materialize_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_materialize_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_materialize_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_materialize_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_materialize_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_materialize_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_materialize_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_materialize_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_materialize_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_materialize_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_materialize_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_materialize_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lazy_clone_view_materialize_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_log_normal_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_log_normal_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_log_normal_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_log_normal_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_logcumsumexp_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_lognormal_kstest_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_bool_tensor_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_bfloat16_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_bfloat16_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_bool_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_bool_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_complex128_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_complex128_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_complex64_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_complex64_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_float16_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_float16_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_float32_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_float32_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_float64_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_float64_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_int16_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_int16_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_int32_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_int32_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_int64_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_int64_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_int8_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_int8_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_uint8_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_cuda_uint8_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_fill_mem_overlap_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_bool_tensor_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_inplace_noncontiguous_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_large_tensor_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_scatter_mem_overlap_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_select_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_select_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_select_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_select_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_select_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_select_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_select_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_select_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_select_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_select_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_select_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_select_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_masked_select_discontiguous_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_memory_format_clone_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_memory_format_consistency_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_memory_format_cpu_and_cuda_ops_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_memory_format_empty_like_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_memory_format_factory_like_functions_preserve_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_memory_format_operators_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_memory_format_preserved_after_permute_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_memory_format_propagation_rules_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_memory_format_to_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_memory_format_type_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_memory_format_type_shortcuts_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_module_share_memory_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_multinomial_cpu_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_multinomial_cpu_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_multinomial_cpu_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_multinomial_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_multinomial_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_multinomial_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_multinomial_deterministic_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_multinomial_deterministic_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_multinomial_deterministic_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_multinomial_device_constrain_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_multinomial_empty_w_replacement_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_multinomial_empty_wo_replacement_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_multinomial_gpu_device_constrain_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_multinomial_rng_state_advance_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_narrow_copy_non_contiguous_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_narrow_empty_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_AdaptiveAvgPool2d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_AdaptiveAvgPool3d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_AdaptiveMaxPool2d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_AvgPool3d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_CTCLoss_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_EmbeddingBag_max_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_FractionalMaxPool2d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_FractionalMaxPool3d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_MaxPool3d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_MaxUnpool1d_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_MaxUnpool1d_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_MaxUnpool1d_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_MaxUnpool2d_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_MaxUnpool2d_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_MaxUnpool2d_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_MaxUnpool3d_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_MaxUnpool3d_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_MaxUnpool3d_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_NLLLoss_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_ReflectionPad1d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_ReflectionPad2d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_ReflectionPad3d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_ReplicationPad1d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_ReplicationPad2d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_ReplicationPad3d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_bincount_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_cumsum_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_cumsum_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_cumsum_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_cumsum_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_cumsum_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_cumsum_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_cumsum_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_cumsum_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_cumsum_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_cumsum_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_grid_sample_2d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_grid_sample_3d_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_histc_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_interpolate_bicubic_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_interpolate_bilinear_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_interpolate_linear_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_interpolate_trilinear_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_kthvalue_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_median_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_put_accumulate_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_alert_put_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_resize_quantized_cuda_qint32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_resize_quantized_cuda_qint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_resize_quantized_cuda_quint2x4, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_resize_quantized_cuda_quint4x2, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nondeterministic_resize_quantized_cuda_quint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_normal_kstest_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_normal_kstest_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_normal_kstest_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_normal_kstest_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_nullary_op_mem_overlap_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_pairwise_distance_empty_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_parallel_cow_materialize_error_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_parallel_cow_materialize_error_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_parallel_cow_materialize_error_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_parallel_cow_materialize_error_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_parallel_cow_materialize_error_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_parallel_cow_materialize_error_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_parallel_cow_materialize_error_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_parallel_cow_materialize_error_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_parallel_cow_materialize_error_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_parallel_cow_materialize_error_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_parallel_cow_materialize_error_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_parallel_cow_materialize_error_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_params_invalidated_with_grads_invalidated_between_unscale_and_step_AdamW_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_params_invalidated_with_grads_invalidated_between_unscale_and_step_Adam_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_params_invalidated_with_grads_invalidated_between_unscale_and_step_SGD_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_pdist_empty_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_pdist_norm_large_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_pickle_gradscaler_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_pin_memory_from_constructor_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_accumulate_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_accumulate_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_accumulate_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_accumulate_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_accumulate_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_accumulate_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_accumulate_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_accumulate_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_accumulate_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_accumulate_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_accumulate_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_empty_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_put_mem_overlap_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_reduced_type_float_copy_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_reduced_type_float_copy_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_repeat_interleave_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scalar_check_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_add_bool_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_add_non_unique_index_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_add_one_dim_deterministic_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_add_to_large_input_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_bool_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_mem_overlap_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_multiply_unsupported_dtypes_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_multiply_unsupported_dtypes_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_non_unique_index_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_non_unique_index_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_non_unique_index_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_non_unique_index_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_non_unique_index_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_non_unique_index_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_non_unique_index_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_non_unique_index_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_non_unique_index_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_non_unique_index_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_non_unique_index_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_non_unique_index_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_operations_to_large_input_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_operations_to_large_input_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_operations_to_large_input_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_operations_to_large_input_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_operations_to_large_input_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_operations_to_large_input_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_operations_to_large_input_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_operations_to_large_input_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_operations_to_large_input_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_operations_to_large_input_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_operations_to_large_input_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_operations_to_large_input_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_scalar_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_scalar_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_scalar_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_scalar_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_scalar_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_scalar_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_scalar_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_scalar_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_scalar_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_scalar_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_scalar_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_reduce_scalar_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_to_large_input_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_scatter_zero_size_index_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_serialization_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_set_default_tensor_type_warnings_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_set_storage_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_set_storage_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_set_storage_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_set_storage_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_set_storage_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_set_storage_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_set_storage_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_set_storage_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_set_storage_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_set_storage_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_set_storage_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_set_storage_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_shift_mem_overlap_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_skip_xla_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_all_devices_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_cuda_uint16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_cuda_uint32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_cuda_uint64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_errors_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_errors_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_errors_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_errors_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_errors_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_errors_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_errors_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_errors_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_errors_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_errors_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_errors_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_errors_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_from_tensor_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_from_tensor_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_from_tensor_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_from_tensor_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_from_tensor_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_from_tensor_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_from_tensor_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_from_tensor_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_from_tensor_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_from_tensor_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_from_tensor_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_from_tensor_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_ok_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_ok_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_ok_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_ok_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_ok_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_ok_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_ok_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_ok_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_ok_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_ok_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_ok_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_meta_ok_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_setitem_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_setitem_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_setitem_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_setitem_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_setitem_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_setitem_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_setitem_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_setitem_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_setitem_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_setitem_cuda_qint32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_setitem_cuda_qint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_setitem_cuda_quint4x2, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_setitem_cuda_quint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_storage_setitem_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_strides_propagation_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_sync_warning_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_take_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_take_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_take_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_take_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_take_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_take_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_take_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_take_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_take_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_take_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_take_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_take_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_take_empty_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_uint16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_uint32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_uint64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_from_storage_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_set_errors_multigpu_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_shape_empty_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_storage_type_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_storage_type_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_storage_type_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_storage_type_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_storage_type_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_storage_type_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_storage_type_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_storage_type_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_storage_type_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_storage_type_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_storage_type_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_storage_type_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_tensor_type_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_ternary_op_mem_overlap_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_typed_storage_meta_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_typed_storage_meta_cuda_bool, test/test_torch.py::TestTorchDeviceTypeCUDA::test_typed_storage_meta_cuda_complex128, test/test_torch.py::TestTorchDeviceTypeCUDA::test_typed_storage_meta_cuda_complex64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_typed_storage_meta_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_typed_storage_meta_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_typed_storage_meta_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_typed_storage_meta_cuda_int16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_typed_storage_meta_cuda_int32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_typed_storage_meta_cuda_int64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_typed_storage_meta_cuda_int8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_typed_storage_meta_cuda_uint8, test/test_torch.py::TestTorchDeviceTypeCUDA::test_unfold_all_devices_and_dtypes_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_unfold_scalars_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_uniform_kstest_cuda_bfloat16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_uniform_kstest_cuda_float16, test/test_torch.py::TestTorchDeviceTypeCUDA::test_uniform_kstest_cuda_float32, test/test_torch.py::TestTorchDeviceTypeCUDA::test_uniform_kstest_cuda_float64, test/test_torch.py::TestTorchDeviceTypeCUDA::test_untyped_storage_meta_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_warn_always_caught_cuda, test/test_torch.py::TestTorchDeviceTypeCUDA::test_where_scalar_handcrafted_values_cuda, test/test_torch.py::TestDevicePrecisionCUDA::test_advancedindex_mixed_cpu_devices_cuda, test/test_torch.py::TestDevicePrecisionCUDA::test_advancedindex_mixed_devices_error_cuda, test/test_torch.py::TestDevicePrecisionCUDA::test_clamp_cuda_float32, test/test_torch.py::TestDevicePrecisionCUDA::test_clamp_cuda_float64, test/test_torch.py::TestDevicePrecisionCUDA::test_clamp_cuda_int64, test/test_torch.py::TestDevicePrecisionCUDA::test_copy_broadcast_cuda, test/test_torch.py::TestDevicePrecisionCUDA::test_copy_noncontig_cuda, test/test_torch.py::TestDevicePrecisionCUDA::test_cuda_device_idx_cuda, test/test_torch.py::TestDevicePrecisionCUDA::test_device_serialization_cuda, test/test_torch.py::TestDevicePrecisionCUDA::test_from_sequence_cuda_float16, test/test_torch.py::TestDevicePrecisionCUDA::test_from_sequence_cuda_float32, test/test_torch.py::TestDevicePrecisionCUDA::test_from_sequence_cuda_float64, test/test_torch.py::TestDevicePrecisionCUDA::test_from_sequence_cuda_int16, test/test_torch.py::TestDevicePrecisionCUDA::test_from_sequence_cuda_int32, test/test_torch.py::TestDevicePrecisionCUDA::test_from_sequence_cuda_int64, test/test_torch.py::TestDevicePrecisionCUDA::test_from_sequence_cuda_int8, test/test_torch.py::TestDevicePrecisionCUDA::test_from_sequence_cuda_uint8, test/test_torch.py::TestDevicePrecisionCUDA::test_index_add_bfloat16_cuda, test/test_torch.py::TestDevicePrecisionCUDA::test_multidevice_serialization_cuda, test/test_torch.py::TestDevicePrecisionCUDA::test_type_conversions_same_device_cuda 2024-08-06T20:51:45.8952695Z 2024-08-06T20:51:45.8953006Z Running test_namedtuple_return_api 1/1 ... [2024-08-06 20:51:45.822719] 2024-08-06T20:51:45.8954439Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_namedtuple_return_api.py', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:51:45.823195] 2024-08-06T20:51:51.2981993Z 2024-08-06T20:51:51.2984236Z test_namedtuple_return_api 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_namedtuple_return_api_1.1_f9107d81858f3a37_.log 2024-08-06T20:51:51.2988366Z Running 3 items in this shard: test/test_namedtuple_return_api.py::TestNamedTupleAPI::test_import_return_types, test/test_namedtuple_return_api.py::TestNamedTupleAPI::test_namedtuple_return, test/test_namedtuple_return_api.py::TestNamedTupleAPI::test_native_functions_yaml 2024-08-06T20:51:51.2991189Z 2024-08-06T20:51:51.2991608Z Running test_decomp 5/24 ... [2024-08-06 20:51:51.298629] 2024-08-06T20:51:51.2997694Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_decomp.py', '-m', 'serial', '--shard-id=5', '--num-shards=24', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:51:51.299308] 2024-08-06T20:51:58.4771154Z 2024-08-06T20:51:58.4773074Z test_decomp 5/24 was successful, full logs can be found in artifacts with path test/test-reports/test_decomp_5.24_c3f5a8e4c029509f_.log 2024-08-06T20:51:58.4774758Z Running 0 items in this shard: 2024-08-06T20:51:58.4775234Z 2024-08-06T20:51:58.4778880Z Running test_decomp 14/24 ... [2024-08-06 20:51:58.477528] 2024-08-06T20:51:58.4785518Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_decomp.py', '-m', 'serial', '--shard-id=14', '--num-shards=24', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:51:58.478092] 2024-08-06T20:52:05.7063791Z 2024-08-06T20:52:05.7064943Z test_decomp 14/24 was successful, full logs can be found in artifacts with path test/test-reports/test_decomp_14.24_e42667637da2d357_.log 2024-08-06T20:52:05.7065791Z Running 0 items in this shard: 2024-08-06T20:52:05.7066057Z 2024-08-06T20:52:05.7069860Z Running test_decomp 15/24 ... [2024-08-06 20:52:05.706645] 2024-08-06T20:52:05.7074878Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_decomp.py', '-m', 'serial', '--shard-id=15', '--num-shards=24', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:52:05.707130] 2024-08-06T20:52:12.9859382Z 2024-08-06T20:52:12.9861376Z test_decomp 15/24 was successful, full logs can be found in artifacts with path test/test-reports/test_decomp_15.24_53521d8677d3baa5_.log 2024-08-06T20:52:12.9863058Z Running 0 items in this shard: 2024-08-06T20:52:12.9863529Z 2024-08-06T20:52:12.9867501Z Running test_decomp 24/24 ... [2024-08-06 20:52:12.986382] 2024-08-06T20:52:12.9874320Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_decomp.py', '-m', 'serial', '--shard-id=24', '--num-shards=24', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:52:12.986958] 2024-08-06T20:52:20.1648837Z 2024-08-06T20:52:20.1649928Z test_decomp 24/24 was successful, full logs can be found in artifacts with path test/test-reports/test_decomp_24.24_4ae095422df96e07_.log 2024-08-06T20:52:20.1651116Z Running 0 items in this shard: 2024-08-06T20:52:20.1651397Z 2024-08-06T20:52:20.1654612Z Running test_model_exports_to_core_aten 1/1 ... [2024-08-06 20:52:20.165142] 2024-08-06T20:52:20.1659667Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_model_exports_to_core_aten.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:52:20.165600] 2024-08-06T20:52:23.9375240Z 2024-08-06T20:52:23.9376410Z test_model_exports_to_core_aten 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_model_exports_to_core_aten_1.1_92e0e483d1c466b3_.log 2024-08-06T20:52:23.9377425Z Running 0 items in this shard: 2024-08-06T20:52:23.9377710Z 2024-08-06T20:52:23.9380969Z Running test_transformers 1/1 ... [2024-08-06 20:52:23.937819] 2024-08-06T20:52:23.9386995Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_transformers.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:52:23.938326] 2024-08-06T20:52:41.2856800Z 2024-08-06T20:52:41.2857788Z test_transformers 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_transformers_1.1_8dcd81afa21e5d8d_.log 2024-08-06T20:52:41.2858694Z Running 0 items in this shard: 2024-08-06T20:52:41.2858960Z 2024-08-06T20:52:41.2862552Z Running functorch/test_ops 8/9 ... [2024-08-06 20:52:41.285953] 2024-08-06T20:52:41.2867744Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'functorch/test_ops.py', '-m', 'serial', '--shard-id=8', '--num-shards=9', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:52:41.286415] 2024-08-06T20:52:49.0654833Z 2024-08-06T20:52:49.0655910Z functorch/test_ops 8/9 was successful, full logs can be found in artifacts with path test/test-reports/functorch.test_ops_8.9_b9165706ceebc288_.log 2024-08-06T20:52:49.0656822Z Running 0 items in this shard: 2024-08-06T20:52:49.0657065Z 2024-08-06T20:52:49.0660490Z Running test_proxy_tensor 1/1 ... [2024-08-06 20:52:49.065761] 2024-08-06T20:52:49.0666208Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_proxy_tensor.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:52:49.066240] 2024-08-06T20:52:54.1401982Z 2024-08-06T20:52:54.1402969Z test_proxy_tensor 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_proxy_tensor_1.1_e6667a18940f47c4_.log 2024-08-06T20:52:54.1403882Z Running 0 items in this shard: 2024-08-06T20:52:54.1404598Z 2024-08-06T20:52:54.1408166Z Running test_foreach 1/1 ... [2024-08-06 20:52:54.140514] 2024-08-06T20:52:54.1413763Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_foreach.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:52:54.140989] 2024-08-06T20:52:59.9161975Z 2024-08-06T20:52:59.9163551Z test_foreach 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_foreach_1.1_0fcd8f2885d33bb7_.log 2024-08-06T20:52:59.9165191Z Running 0 items in this shard: 2024-08-06T20:52:59.9165650Z 2024-08-06T20:52:59.9167976Z Running test_ops_gradients 1/4 ... [2024-08-06 20:52:59.916417] 2024-08-06T20:52:59.9172690Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_ops_gradients.py', '-m', 'serial', '--shard-id=1', '--num-shards=4', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:52:59.916907] 2024-08-06T20:53:06.4932835Z 2024-08-06T20:53:06.4934368Z test_ops_gradients 1/4 was successful, full logs can be found in artifacts with path test/test-reports/test_ops_gradients_1.4_a9ece66daa539b1c_.log 2024-08-06T20:53:06.4935289Z Running 0 items in this shard: 2024-08-06T20:53:06.4935557Z 2024-08-06T20:53:06.4938368Z Running test_meta 5/5 ... [2024-08-06 20:53:06.493534] 2024-08-06T20:53:06.4943593Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_meta.py', '-m', 'serial', '--shard-id=5', '--num-shards=5', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:53:06.494016] 2024-08-06T20:53:21.2859803Z 2024-08-06T20:53:21.2860895Z test_meta 5/5 was successful, full logs can be found in artifacts with path test/test-reports/test_meta_5.5_ec05767ff76bc58d_.log 2024-08-06T20:53:21.2861734Z Running 0 items in this shard: 2024-08-06T20:53:21.2861999Z 2024-08-06T20:53:21.2865453Z Running test_ops 2/11 ... [2024-08-06 20:53:21.286258] 2024-08-06T20:53:21.2870583Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_ops.py', '-m', 'serial', '--shard-id=2', '--num-shards=11', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:53:21.286729] 2024-08-06T20:53:34.9275977Z 2024-08-06T20:53:34.9277091Z test_ops 2/11 was successful, full logs can be found in artifacts with path test/test-reports/test_ops_2.11_dc32d920906cf18b_.log 2024-08-06T20:53:34.9277886Z Running 0 items in this shard: 2024-08-06T20:53:34.9278151Z 2024-08-06T20:53:34.9280841Z Running test_ops 9/11 ... [2024-08-06 20:53:34.927820] 2024-08-06T20:53:34.9285764Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_ops.py', '-m', 'serial', '--shard-id=9', '--num-shards=11', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:53:34.928256] 2024-08-06T20:53:48.8687133Z 2024-08-06T20:53:48.8688052Z test_ops 9/11 was successful, full logs can be found in artifacts with path test/test-reports/test_ops_9.11_0a0a65c8b31e19fe_.log 2024-08-06T20:53:48.8688864Z Running 0 items in this shard: 2024-08-06T20:53:48.8689108Z 2024-08-06T20:53:48.8693311Z Running test_module_tracker 1/1 ... [2024-08-06 20:53:48.869026] 2024-08-06T20:53:48.8698835Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_module_tracker.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:53:48.869524] 2024-08-06T20:53:52.2907771Z 2024-08-06T20:53:52.2908738Z test_module_tracker 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_module_tracker_1.1_e6469be478c2e430_.log 2024-08-06T20:53:52.2909689Z Running 0 items in this shard: 2024-08-06T20:53:52.2909932Z 2024-08-06T20:53:52.2913986Z Running test_comparison_utils 1/1 ... [2024-08-06 20:53:52.291090] 2024-08-06T20:53:52.2919829Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_comparison_utils.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:53:52.291598] 2024-08-06T20:53:55.6626826Z 2024-08-06T20:53:55.6628060Z test_comparison_utils 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_comparison_utils_1.1_ec0fbb2788e0faa4_.log 2024-08-06T20:53:55.6629004Z Running 0 items in this shard: 2024-08-06T20:53:55.6629251Z 2024-08-06T20:53:55.6632250Z Running functorch/test_vmap 1/1 ... [2024-08-06 20:53:55.662929] 2024-08-06T20:53:55.6637561Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'functorch/test_vmap.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:53:55.663388] 2024-08-06T20:54:01.2882596Z 2024-08-06T20:54:01.2883768Z functorch/test_vmap 1/1 was successful, full logs can be found in artifacts with path test/test-reports/functorch.test_vmap_1.1_c8522dc184c01937_.log 2024-08-06T20:54:01.2884694Z Running 0 items in this shard: 2024-08-06T20:54:01.2884944Z 2024-08-06T20:54:01.2888927Z Running inductor/test_inplacing_pass 1/1 ... [2024-08-06 20:54:01.288509] 2024-08-06T20:54:01.2893300Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_inplacing_pass.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:54:01.288964] 2024-08-06T20:54:04.6291494Z 2024-08-06T20:54:04.6292840Z inductor/test_inplacing_pass 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_inplacing_pass_1.1_8e394ec6337397f3_.log 2024-08-06T20:54:04.6293757Z 2024-08-06T20:54:04.6301034Z Running export/test_tools 1/1 ... [2024-08-06 20:54:04.629794] 2024-08-06T20:54:04.6308806Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'export/test_tools.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:54:04.630502] 2024-08-06T20:54:08.6032896Z 2024-08-06T20:54:08.6034081Z export/test_tools 1/1 was successful, full logs can be found in artifacts with path test/test-reports/export.test_tools_1.1_e61989030b35495d_.log 2024-08-06T20:54:08.6034979Z Running 0 items in this shard: 2024-08-06T20:54:08.6035227Z 2024-08-06T20:54:08.6038530Z Running inductor/test_torchinductor_dynamic_shapes 1/4 ... [2024-08-06 20:54:08.603547] 2024-08-06T20:54:08.6043692Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_torchinductor_dynamic_shapes.py', '-m', 'serial', '--shard-id=1', '--num-shards=4', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:54:08.604005] 2024-08-06T20:54:18.3874511Z 2024-08-06T20:54:18.3876040Z inductor/test_torchinductor_dynamic_shapes 1/4 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_torchinductor_dynamic_shapes_1.4_5bea70ed43c933aa_.log 2024-08-06T20:54:18.3878138Z Running 1 items in this shard: test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_large_block_sizes_dynamic_shapes_cpu 2024-08-06T20:54:18.3878930Z 2024-08-06T20:54:18.3879235Z Running torch_np/test_binary_ufuncs 1/1 ... [2024-08-06 20:54:18.387503] 2024-08-06T20:54:18.3883467Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'torch_np/test_binary_ufuncs.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:54:18.387987] 2024-08-06T20:54:21.8595691Z 2024-08-06T20:54:21.8596864Z torch_np/test_binary_ufuncs 1/1 was successful, full logs can be found in artifacts with path test/test-reports/torch_np.test_binary_ufuncs_1.1_982b23dcb1a41e4e_.log 2024-08-06T20:54:21.8597844Z Running 0 items in this shard: 2024-08-06T20:54:21.8598109Z 2024-08-06T20:54:21.8602338Z Running dynamo/test_inline_inbuilt_nn_modules 1/1 ... [2024-08-06 20:54:21.859902] 2024-08-06T20:54:21.8608900Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'dynamo/test_inline_inbuilt_nn_modules.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:54:21.860467] 2024-08-06T20:54:27.9370767Z 2024-08-06T20:54:27.9372960Z dynamo/test_inline_inbuilt_nn_modules 1/1 was successful, full logs can be found in artifacts with path test/test-reports/dynamo.test_inline_inbuilt_nn_modules_1.1_c19a33c832a18dcf_.log 2024-08-06T20:54:27.9375828Z Running 0 items in this shard: 2024-08-06T20:54:27.9376307Z 2024-08-06T20:54:27.9379347Z Running inductor/test_inductor_utils 1/1 ... [2024-08-06 20:54:27.937552] 2024-08-06T20:54:27.9386612Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_inductor_utils.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:54:27.938183] 2024-08-06T20:54:31.3598049Z 2024-08-06T20:54:31.3599688Z inductor/test_inductor_utils 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_inductor_utils_1.1_6685839cf3a248f3_.log 2024-08-06T20:54:31.3600688Z Running 0 items in this shard: 2024-08-06T20:54:31.3602010Z 2024-08-06T20:54:31.3605494Z Running torch_np/test_unary_ufuncs 1/1 ... [2024-08-06 20:54:31.360175] 2024-08-06T20:54:31.3610474Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'torch_np/test_unary_ufuncs.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:54:31.360658] 2024-08-06T20:54:34.7817273Z 2024-08-06T20:54:34.7819359Z torch_np/test_unary_ufuncs 1/1 was successful, full logs can be found in artifacts with path test/test-reports/torch_np.test_unary_ufuncs_1.1_ca25b089fa697dca_.log 2024-08-06T20:54:34.7821260Z Running 0 items in this shard: 2024-08-06T20:54:34.7821713Z 2024-08-06T20:54:34.7824515Z Running inductor/test_extension_backend 1/1 ... [2024-08-06 20:54:34.782096] 2024-08-06T20:54:34.7832069Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_extension_backend.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:54:34.782732] 2024-08-06T20:54:41.9611087Z 2024-08-06T20:54:41.9613129Z inductor/test_extension_backend 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_extension_backend_1.1_92556c499ca43ed7_.log 2024-08-06T20:54:41.9615225Z Running 0 items in this shard: 2024-08-06T20:54:41.9618893Z 2024-08-06T20:54:41.9619534Z Running export/test_export_nonstrict 1/1 ... [2024-08-06 20:54:41.961573] 2024-08-06T20:54:41.9626309Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'export/test_export_nonstrict.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:54:41.962154] 2024-08-06T20:54:47.5374334Z 2024-08-06T20:54:47.5375639Z export/test_export_nonstrict 1/1 was successful, full logs can be found in artifacts with path test/test-reports/export.test_export_nonstrict_1.1_6fbca26bceb97d5c_.log 2024-08-06T20:54:47.5376698Z Running 0 items in this shard: 2024-08-06T20:54:47.5376992Z 2024-08-06T20:54:47.5379268Z Running inductor/test_triton_extension_backend 1/1 ... [2024-08-06 20:54:47.537630] 2024-08-06T20:54:47.5384074Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_triton_extension_backend.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:54:47.538066] 2024-08-06T20:54:54.7677501Z 2024-08-06T20:54:54.7678730Z inductor/test_triton_extension_backend 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_triton_extension_backend_1.1_00cac446b197b5d7_.log 2024-08-06T20:54:54.7679845Z Running 0 items in this shard: 2024-08-06T20:54:54.7680089Z 2024-08-06T20:54:54.7686947Z Running inductor/test_triton_kernels 1/1 ... [2024-08-06 20:54:54.768404] 2024-08-06T20:54:54.7694104Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_triton_kernels.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:54:54.769045] 2024-08-06T20:54:58.3912529Z 2024-08-06T20:54:58.3913970Z inductor/test_triton_kernels 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_triton_kernels_1.1_05e9541c9b0db076_.log 2024-08-06T20:54:58.3916647Z Running 0 items in this shard: 2024-08-06T20:54:58.3916940Z 2024-08-06T20:54:58.3919014Z Running dynamo/test_base_output 1/1 ... [2024-08-06 20:54:58.391571] 2024-08-06T20:54:58.3924704Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'dynamo/test_base_output.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:54:58.392071] 2024-08-06T20:55:01.9135964Z 2024-08-06T20:55:01.9138037Z dynamo/test_base_output 1/1 was successful, full logs can be found in artifacts with path test/test-reports/dynamo.test_base_output_1.1_5f87dc0ecad03c02_.log 2024-08-06T20:55:01.9139948Z Running 0 items in this shard: 2024-08-06T20:55:01.9140397Z 2024-08-06T20:55:01.9141357Z Running export/test_serdes 1/1 ... [2024-08-06 20:55:01.913812] 2024-08-06T20:55:01.9146848Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'export/test_serdes.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:55:01.914308] 2024-08-06T20:55:07.4894309Z 2024-08-06T20:55:07.4895321Z export/test_serdes 1/1 was successful, full logs can be found in artifacts with path test/test-reports/export.test_serdes_1.1_154a92ee259f8afe_.log 2024-08-06T20:55:07.4896259Z Running 0 items in this shard: 2024-08-06T20:55:07.4896503Z 2024-08-06T20:55:07.4899280Z Running dynamo/test_interop 1/1 ... [2024-08-06 20:55:07.489655] 2024-08-06T20:55:07.4904577Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'dynamo/test_interop.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:55:07.490075] 2024-08-06T20:55:10.9110172Z 2024-08-06T20:55:10.9111371Z dynamo/test_interop 1/1 was successful, full logs can be found in artifacts with path test/test-reports/dynamo.test_interop_1.1_d1ce67b3889c666f_.log 2024-08-06T20:55:10.9112298Z Running 0 items in this shard: 2024-08-06T20:55:10.9112549Z 2024-08-06T20:55:10.9115569Z Running inductor/test_compile_worker 1/1 ... [2024-08-06 20:55:10.911247] 2024-08-06T20:55:10.9120152Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_compile_worker.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:55:10.911668] 2024-08-06T20:55:17.1378951Z 2024-08-06T20:55:17.1380162Z inductor/test_compile_worker 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_compile_worker_1.1_1ba16cb14fd7b18d_.log 2024-08-06T20:55:17.1381183Z Running 0 items in this shard: 2024-08-06T20:55:17.1381453Z 2024-08-06T20:55:17.1384132Z Running dynamo/test_fx_passes_pre_grad 1/1 ... [2024-08-06 20:55:17.138101] 2024-08-06T20:55:17.1388764Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'dynamo/test_fx_passes_pre_grad.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:55:17.138533] 2024-08-06T20:55:20.6599255Z 2024-08-06T20:55:20.6600744Z dynamo/test_fx_passes_pre_grad 1/1 was successful, full logs can be found in artifacts with path test/test-reports/dynamo.test_fx_passes_pre_grad_1.1_4ed62af637d74e6c_.log 2024-08-06T20:55:20.6601917Z Running 0 items in this shard: 2024-08-06T20:55:20.6602185Z 2024-08-06T20:55:20.6604559Z Running inductor/test_flex_decoding 1/1 ... [2024-08-06 20:55:20.660154] 2024-08-06T20:55:20.6610062Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_flex_decoding.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:55:20.660654] 2024-08-06T20:55:24.3823867Z 2024-08-06T20:55:24.3825192Z inductor/test_flex_decoding 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_flex_decoding_1.1_b9cd1c4c95d146cc_.log 2024-08-06T20:55:24.3826182Z Running 0 items in this shard: 2024-08-06T20:55:24.3826451Z 2024-08-06T20:55:24.3828413Z Running dynamo/test_skip_non_tensor 1/1 ... [2024-08-06 20:55:24.382548] 2024-08-06T20:55:24.3833488Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'dynamo/test_skip_non_tensor.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:55:24.383008] 2024-08-06T20:55:27.7538043Z 2024-08-06T20:55:27.7539377Z dynamo/test_skip_non_tensor 1/1 was successful, full logs can be found in artifacts with path test/test-reports/dynamo.test_skip_non_tensor_1.1_8d8d7967a644ae3f_.log 2024-08-06T20:55:27.7540377Z Running 0 items in this shard: 2024-08-06T20:55:27.7540629Z 2024-08-06T20:55:27.7543615Z Running inductor/test_move_constructors_to_cuda 1/1 ... [2024-08-06 20:55:27.754026] 2024-08-06T20:55:27.7548612Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_move_constructors_to_cuda.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:55:27.754490] 2024-08-06T20:55:31.3620425Z 2024-08-06T20:55:31.3621804Z inductor/test_move_constructors_to_cuda 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_move_constructors_to_cuda_1.1_9dcb182ea2d56995_.log 2024-08-06T20:55:31.3622783Z 2024-08-06T20:55:31.3625819Z Running test_matmul_cuda 1/1 ... [2024-08-06 20:55:31.362265] 2024-08-06T20:55:31.3631737Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_matmul_cuda.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:55:31.362760] 2024-08-06T20:55:35.4854496Z 2024-08-06T20:55:35.4855845Z test_matmul_cuda 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_matmul_cuda_1.1_28e8348588e14723_.log 2024-08-06T20:55:35.4856746Z Running 0 items in this shard: 2024-08-06T20:55:35.4857012Z 2024-08-06T20:55:35.4860713Z Running dynamo/test_sdpa 1/1 ... [2024-08-06 20:55:35.485754] 2024-08-06T20:55:35.4866032Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'dynamo/test_sdpa.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:55:35.486226] 2024-08-06T20:55:38.9074044Z 2024-08-06T20:55:38.9075239Z dynamo/test_sdpa 1/1 was successful, full logs can be found in artifacts with path test/test-reports/dynamo.test_sdpa_1.1_e6e2d1c104b92330_.log 2024-08-06T20:55:38.9076137Z Running 0 items in this shard: 2024-08-06T20:55:38.9076389Z 2024-08-06T20:55:38.9079875Z Running inductor/test_metrics 1/1 ... [2024-08-06 20:55:38.907668] 2024-08-06T20:55:38.9087304Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_metrics.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:55:38.908172] 2024-08-06T20:55:42.5799261Z 2024-08-06T20:55:42.5800457Z inductor/test_metrics 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_metrics_1.1_59af11765b8ea9e9_.log 2024-08-06T20:55:42.5801270Z 2024-08-06T20:55:42.5805179Z Running inductor/test_debug_trace 1/1 ... [2024-08-06 20:55:42.580171] 2024-08-06T20:55:42.5810859Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_debug_trace.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:55:42.580701] 2024-08-06T20:55:50.1600492Z 2024-08-06T20:55:50.1601872Z inductor/test_debug_trace 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_debug_trace_1.1_c75847ce10512bcf_.log 2024-08-06T20:55:50.1602895Z Running 0 items in this shard: 2024-08-06T20:55:50.1603143Z 2024-08-06T20:55:50.1606216Z Running test_modules 1/2 ... [2024-08-06 20:55:50.160262] 2024-08-06T20:55:50.1611182Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_modules.py', '-m', 'serial', '--shard-id=1', '--num-shards=2', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:55:50.160743] 2024-08-06T20:55:56.0362463Z 2024-08-06T20:55:56.0363597Z test_modules 1/2 was successful, full logs can be found in artifacts with path test/test-reports/test_modules_1.2_fa9b04ae345eaf52_.log 2024-08-06T20:55:56.0364628Z Running 0 items in this shard: 2024-08-06T20:55:56.0364965Z 2024-08-06T20:55:56.0367760Z Running inductor/test_cuda_cpp_wrapper 1/1 ... [2024-08-06 20:55:56.036473] 2024-08-06T20:55:56.0372876Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_cuda_cpp_wrapper.py', '-m', 'serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:55:56.036901] 2024-08-06T20:56:03.4903639Z 2024-08-06T20:56:03.4905302Z inductor/test_cuda_cpp_wrapper 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_cuda_cpp_wrapper_1.1_d07af4907c36874a_.log 2024-08-06T20:56:03.4906212Z 2024-08-06T20:56:03.4966003Z Running test_decomp 5/24 ... [2024-08-06 20:56:03.496273] 2024-08-06T20:56:03.4973155Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_decomp.py', '-m', 'not serial', '--shard-id=5', '--num-shards=24', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:56:03.496943] 2024-08-06T20:56:03.5019208Z Running test_decomp 14/24 ... [2024-08-06 20:56:03.501604] 2024-08-06T20:56:03.5027618Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_decomp.py', '-m', 'not serial', '--shard-id=14', '--num-shards=24', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 20:56:03.502381] 2024-08-06T21:03:20.2724365Z 2024-08-06T21:03:20.2725509Z test_decomp 5/24 was successful, full logs can be found in artifacts with path test/test-reports/test_decomp_5.24_189b6118962869f4_.log 2024-08-06T21:03:20.2876098Z Running 405 items in this shard: test/test_decomp.py::TestDecompCUDA::test_comprehensive_T_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive___rdiv___cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive___rmod___cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive___rmul___cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive___rpow___cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive__chunk_cat_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive__unsafe_masked_index_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_acos_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_acos_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_add_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_addbmm_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_addcdiv_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_addr_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_addr_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_alias_copy_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_all_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_all_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_amax_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_aminmax_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_any_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_argmin_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_argsort_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_as_strided_copy_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_as_strided_partial_views_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_asin_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_asin_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_atanh_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_atleast_1d_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_atleast_1d_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_atleast_2d_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_atleast_2d_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_bitwise_and_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_broadcast_to_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_broadcast_to_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_byte_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_byte_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_byte_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_ceil_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_chalf_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_char_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_clamp_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_clamp_max_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_clone_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_column_stack_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_combinations_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_combinations_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_contiguous_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cos_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cosh_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_count_nonzero_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_count_nonzero_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cummin_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cummin_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cumprod_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cumprod_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_deg2rad_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diag_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diagonal_copy_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diagonal_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diagonal_scatter_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diagonal_scatter_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_div_no_rounding_mode_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_div_no_rounding_mode_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_dot_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_dot_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_eq_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_equal_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_erf_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_erf_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_erfc_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_exp_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_expand_as_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_expand_as_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_expand_copy_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_fft2_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_ifftn_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_ihfft2_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_irfft2_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_irfft_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_irfftn_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_rfft_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_rfftn_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_flatten_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_flip_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fliplr_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fliplr_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_float_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_full_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_full_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_full_like_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_full_like_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_geqrf_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_gt_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_half_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_half_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_i0_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_index_add_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_index_reduce_mean_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_isnan_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_isreal_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_item_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_jiterator_binary_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_jiterator_binary_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_jiterator_binary_return_by_ref_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_jiterator_binary_return_by_ref_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_jiterator_unary_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_jiterator_unary_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_kthvalue_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_ldexp_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_le_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_lgamma_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_cross_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_diagonal_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_eigh_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_eigh_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_eigvalsh_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_inv_ex_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_ldl_factor_ex_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_norm_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_norm_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_pinv_singular_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_solve_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_vander_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_vander_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_vecdot_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_vector_norm_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linspace_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linspace_tensor_overload_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_log1p_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logical_or_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_long_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_lt_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mT_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mT_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_amax_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_argmax_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_logsumexp_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_prod_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_std_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_sum_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_matrix_exp_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_max_reduction_no_dim_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_max_reduction_with_dim_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_maximum_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_median_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_meshgrid_variadic_tensors_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_min_binary_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_min_reduction_with_dim_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_minimum_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_msort_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mul_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mul_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mvlgamma_mvlgamma_p_1_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_narrow_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_narrow_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_neg_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_new_empty_strided_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_new_zeros_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_new_zeros_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_new_zeros_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_avg_pool1d_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_batch_norm_without_cudnn_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_binary_cross_entropy_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_celu_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_feature_alpha_dropout_without_train_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_hardsigmoid_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_instance_norm_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_interpolate_nearest-exact_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_margin_ranking_loss_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_max_unpool2d_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pad_constant_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pad_constant_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pad_replicate_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pairwise_distance_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_rms_norm_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_softmin_with_dtype_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_softmin_with_dtype_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_softplus_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_softsign_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_softsign_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_threshold_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_norm_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_norm_fro_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_norm_inf_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_outer_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_permute_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_polygamma_polygamma_n_0_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_pow_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_prod_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_quantile_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_real_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_renorm_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_repeat_interleave_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_reshape_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_reshape_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_resize__cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_resize_as__cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_resolve_conj_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_resolve_conj_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_resolve_neg_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_roll_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_rot90_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_rsqrt_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_rsqrt_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_rsub_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scatter_reduce_amin_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scatter_reduce_sum_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sgn_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sigmoid_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sigmoid_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_signbit_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_signbit_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sinc_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sinc_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_airy_ai_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_bessel_j1_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_bessel_j1_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_bessel_y0_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_entr_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_i1_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_modified_bessel_i0_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_shifted_chebyshev_polynomial_t_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_shifted_chebyshev_polynomial_u_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_shifted_chebyshev_polynomial_w_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_zeta_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_split_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_split_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_split_with_sizes_copy_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_squeeze_multiple_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_stack_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_std_unbiased_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sum_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sum_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sum_to_size_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_t_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_take_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_tanh_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_tensor_split_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_tile_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_to_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_to_sparse_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_to_sparse_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_trace_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_transpose_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_transpose_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_trunc_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unbind_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unflatten_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unfold_copy_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unfold_copy_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unsafe_split_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unsafe_split_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unsqueeze_copy_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_var_mean_unbiased_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_view_as_complex_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_view_as_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_view_as_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_zero__cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_zeros_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_zeros_like_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick__batch_norm_with_update_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick__chunk_cat_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick__softmax_backward_data_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_acos_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_acosh_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_add_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_add_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_addcdiv_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_atanh_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_atanh_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_clamp_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_constant_pad_nd_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_block_diag_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_index_copy_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_logaddexp2_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_renorm_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_special_xlog1py_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_t_copy_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_zero__cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_count_nonzero_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_cumprod_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_cumprod_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_deg2rad_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_diag_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_diagonal_scatter_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_digamma_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_div_floor_rounding_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_div_trunc_rounding_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_dot_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_empty_like_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_empty_strided_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_empty_strided_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_erf_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_erfc_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_erfc_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_erfinv_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_erfinv_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_exp_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_exp_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_expand_copy_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_expand_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_eye_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_fft_hfft2_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_fft_hfft_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_fft_hfft_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_fft_ifft2_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_fft_ifft_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_fft_irfft_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_fft_irfftn_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_fft_rfft2_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_fft_rfft2_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_fft_rfftn_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_fmod_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_ge_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_geometric_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_geometric_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_gt_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_i0_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_i0_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_index_add_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_quick_index_copy_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_index_fill_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_index_fill_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_index_select_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_index_select_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_isinf_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_isposinf_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_lcm_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_lerp_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_linalg_vector_norm_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_log10_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_log_normal_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_logical_and_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_logical_xor_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_logit_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_logsumexp_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_masked_fill_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_meshgrid_variadic_tensors_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_meshgrid_variadic_tensors_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_minimum_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_minimum_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_mvlgamma_mvlgamma_p_3_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_nan_to_num_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_new_empty_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_new_full_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_new_full_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_binary_cross_entropy_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_binary_cross_entropy_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_glu_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_hardshrink_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_mish_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_mse_loss_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_pad_constant_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_pad_constant_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_prelu_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_silu_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_unfold_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_norm_nuc_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_ones_like_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_quick_permute_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_permute_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_prod_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_reciprocal_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_repeat_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_roll_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_rot90_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_round_decimals_neg_3_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_select_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_quick_select_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_sgn_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_sgn_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_sin_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_sinh_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_slice_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_slice_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_special_ndtri_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_special_ndtri_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_special_xlog1py_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_special_xlog1py_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_special_zeta_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_split_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_split_with_sizes_copy_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_split_with_sizes_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_sqrt_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_sqrt_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_squeeze_multiple_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_std_mean_unbiased_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_sum_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_t_copy_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_t_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_t_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_take_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_take_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_triu_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_triu_indices_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_unfold_copy_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_unfold_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_unfold_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_unsqueeze_copy_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_unsqueeze_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_var_mean_unbiased_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_xlogy_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_zero__cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_rnn_decomp_module_nn_RNN_train_mode_cuda_float64, test/test_decomp.py::DecompOneOffTestsCUDA::test_contiguous_log_softmax_cuda 2024-08-06T21:03:20.3021900Z 2024-08-06T21:03:23.9080927Z Running test_decomp 15/24 ... [2024-08-06 21:03:23.907534] 2024-08-06T21:03:23.9084757Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_decomp.py', '-m', 'not serial', '--shard-id=15', '--num-shards=24', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:03:23.907986] 2024-08-06T21:05:26.2814882Z 2024-08-06T21:05:26.2815845Z test_decomp 14/24 was successful, full logs can be found in artifacts with path test/test-reports/test_decomp_14.24_2c4b0a38c842b87b_.log 2024-08-06T21:05:26.2956019Z Running 375 items in this shard: test/test_decomp.py::TestDecompCUDA::test_comprehensive_H_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_H_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_H_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_T_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_T_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_T_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive___radd___cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive___rxor___cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_abs_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_acos_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_addcdiv_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_addcmul_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_addmm_decomposed_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_addmv_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_addr_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_allclose_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_arange_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_arange_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_arange_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_argsort_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_as_strided_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_as_strided_partial_views_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_asin_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_atleast_2d_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_baddbmm_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_bfloat16_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_bool_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_broadcast_to_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_byte_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_byte_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cartesian_prod_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cauchy_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_ceil_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cfloat_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_chalf_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_char_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_chunk_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_clamp_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_clamp_min_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_clone_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_conj_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_conj_physical_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_constant_pad_nd_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_corrcoef_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cos_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cos_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cosh_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cummin_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cummin_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cumprod_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cumulative_trapezoid_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diag_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diag_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diag_embed_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diagonal_scatter_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_div_floor_rounding_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_empty_permuted_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_empty_permuted_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_empty_permuted_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_empty_strided_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_eye_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_fftn_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_fftn_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_fftshift_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_fftshift_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_hfft2_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_ifft2_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_ifft_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_flipud_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_float_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_floor_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_floor_divide_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_floor_divide_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_floor_divide_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fmod_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fmod_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_gather_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_gradient_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_hstack_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_index_add_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_index_put_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_index_reduce_amax_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_int_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_isclose_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_isfinite_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_isinf_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_isneginf_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_isreal_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_jiterator_binary_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_jiterator_binary_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_kron_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_le_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_det_singular_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_diagonal_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_diagonal_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_lu_solve_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_lu_solve_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_matrix_norm_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_matrix_norm_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_svd_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linspace_tensor_overload_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linspace_tensor_overload_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_log10_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_log1p_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_log1p_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_log1p_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_log_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logcumsumexp_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logical_not_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logical_or_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logical_or_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logit_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logit_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_long_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mT_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_cumprod_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_cumsum_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_cumsum_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_prod_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_select_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_sum_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_var_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_matmul_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_max_pool2d_with_indices_backward_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_max_reduction_with_dim_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mean_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_meshgrid_list_of_tensors_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_min_reduction_with_dim_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_movedim_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_msort_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mvlgamma_mvlgamma_p_5_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nanmedian_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_narrow_copy_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_neg_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_new_empty_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_new_empty_strided_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_adaptive_avg_pool1d_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_adaptive_avg_pool2d_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_adaptive_max_pool3d_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_conv3d_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_cosine_embedding_loss_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_elu_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_feature_alpha_dropout_with_train_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_feature_alpha_dropout_without_train_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_fractional_max_pool3d_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_group_norm_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_leaky_relu_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_margin_ranking_loss_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_margin_ranking_loss_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_max_pool1d_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_max_unpool2d_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_multilabel_margin_loss_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_nll_loss_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pad_replicate_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pixel_shuffle_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pixel_unshuffle_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_relu_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_selu_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_smooth_l1_loss_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_softsign_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_triplet_margin_with_distance_loss_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_unfold_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nonzero_static_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_normal_in_place_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_ones_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_ones_like_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_permute_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_polar_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_polygamma_polygamma_n_3_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_polygamma_polygamma_n_4_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_positive_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_put_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_rand_like_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_remainder_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_resize_as__cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_resolve_conj_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_roll_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_rsub_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scalar_tensor_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scalar_tensor_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scatter_reduce_amax_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scatter_reduce_amin_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scatter_reduce_prod_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_select_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_select_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_select_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sgn_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_short_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sigmoid_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sigmoid_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sign_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_signbit_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sinc_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sinh_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sinh_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_slice_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_bessel_j0_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_bessel_j0_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_bessel_j0_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_bessel_j1_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_chebyshev_polynomial_w_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_erfcx_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_erfcx_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_legendre_polynomial_p_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_ndtr_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_polygamma_special_polygamma_n_0_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sqrt_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_stack_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_std_mean_unbiased_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sum_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_t_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_to_sparse_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_torch_ops_aten__safe_softmax_default_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_torch_ops_aten__safe_softmax_default_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_transpose_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_trapezoid_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_triu_indices_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_triu_indices_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_true_divide_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_true_divide_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unbind_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unflatten_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unflatten_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unfold_copy_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unfold_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unravel_index_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unsafe_chunk_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unsafe_split_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_var_mean_unbiased_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_var_mean_unbiased_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_view_copy_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_vsplit_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_vsplit_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_zeros_like_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick__chunk_cat_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick__chunk_cat_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick__unsafe_masked_index_put_accumulate_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick__unsafe_masked_index_put_accumulate_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick__upsample_bilinear2d_aa_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_abs_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_abs_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_abs_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_acos_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_addcmul_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_all_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_any_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_asin_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_atan_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_quick_bitwise_and_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_bitwise_or_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_cat_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_ceil_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_clamp_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_clamp_max_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_clamp_min_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_clone_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_logsumexp_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_norm_nuc_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_squeeze_multiple_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_cos_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_count_nonzero_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_deg2rad_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_diagonal_scatter_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_digamma_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_dist_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_div_no_rounding_mode_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_empty_strided_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_exp2_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_exp2_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_expand_copy_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_eye_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_fft_fft2_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_fft_fft_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_fft_fftn_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_fft_ifft2_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_fft_ifft_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_fft_ifftn_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_fft_ihfftn_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_fft_irfft2_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_fft_irfft2_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_fft_irfftn_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_fft_rfft2_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_fill_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_floor_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_fmax_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_fmod_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_fmod_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_fmod_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_frac_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_gt_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_isin_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_isinf_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_isnan_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_isposinf_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_isposinf_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_item_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_item_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_item_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_le_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_linalg_diagonal_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_log10_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_log2_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_log_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_logical_and_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_logical_not_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_logical_not_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_masked_fill_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_meshgrid_list_of_tensors_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_minimum_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_minimum_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_mul_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_nan_to_num_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_nansum_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_narrow_copy_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_narrow_copy_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_native_batch_norm_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_native_dropout_backward_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_native_layer_norm_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_ne_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_new_empty_strided_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_new_zeros_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_embedding_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_pad_constant_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_prelu_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_prelu_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_relu6_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_softshrink_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_ones_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_ones_like_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_repeat_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_repeat_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_roll_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_rot90_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_select_scatter_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_sigmoid_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_sin_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_sin_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_sinh_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_special_i1_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_special_i1_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_special_ndtr_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_special_ndtr_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_special_ndtr_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_special_xlog1py_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_special_zeta_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_split_list_args_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_split_with_sizes_copy_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_split_with_sizes_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_squeeze_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_quick_squeeze_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_squeeze_multiple_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_squeeze_multiple_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_std_mean_unbiased_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_sub_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_quick_sub_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_sum_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_tanh_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_trace_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_trace_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_transpose_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_quick_unbind_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_unfold_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_unfold_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_unsqueeze_copy_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_var_mean_unbiased_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_vdot_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_vdot_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_where_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_where_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_quick_xlogy_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_zero__cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_zeros_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_zeros_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_uniform_cuda, test/test_decomp.py::HasDecompTest::test_has_decomposition 2024-08-06T21:05:26.3089686Z 2024-08-06T21:05:29.8848007Z Running test_decomp 24/24 ... [2024-08-06 21:05:29.884339] 2024-08-06T21:05:29.8851931Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_decomp.py', '-m', 'not serial', '--shard-id=24', '--num-shards=24', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:05:29.884792] 2024-08-06T21:11:09.8657596Z 2024-08-06T21:11:09.8659268Z test_decomp 15/24 was successful, full logs can be found in artifacts with path test/test-reports/test_decomp_15.24_de5e9d44f2bd194a_.log 2024-08-06T21:11:09.8789134Z Running 351 items in this shard: test/test_decomp.py::TestDecompCUDA::test_comprehensive___rmul___cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive___rmul___cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive___ror___cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive___rsub___cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive__chunk_cat_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_abs_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_acosh_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_addcmul_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_all_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_aminmax_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_aminmax_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_angle_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_arange_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_argsort_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_as_strided_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_as_strided_partial_views_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_atan2_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_atan_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_atleast_1d_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_atleast_3d_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_bernoulli_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_bfloat16_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_bitwise_not_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_bitwise_not_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_broadcast_tensors_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_char_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_combinations_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_conj_physical_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cosh_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_count_nonzero_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cross_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cumprod_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cumsum_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diagflat_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diff_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_digamma_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_div_floor_rounding_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_div_no_rounding_mode_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_dstack_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_empty_like_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_equal_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_fft2_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_fft2_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_fftshift_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_fftshift_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_hfft2_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_hfftn_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_ifft2_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_ifftshift_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_ifftshift_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_ihfft2_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_irfft2_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_irfftn_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_irfftn_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_rfft_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fill_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_flatten_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fliplr_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fliplr_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_flipud_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_float_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_float_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_float_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_floor_divide_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_floor_divide_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_floor_divide_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_floor_divide_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fmax_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fmax_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_full_like_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_gather_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_ge_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_gt_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_gt_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_heaviside_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_index_put_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_index_reduce_mean_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_isclose_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_isfinite_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_isposinf_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_isreal_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_jiterator_unary_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_kron_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_kron_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_ldexp_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_le_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_cross_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_diagonal_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_eig_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_ldl_factor_ex_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_lstsq_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_lu_solve_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_tensorsolve_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linspace_tensor_overload_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_log10_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_log_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_log_normal_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logcumsumexp_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logspace_tensor_overload_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_lt_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_lu_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mT_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_cumprod_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_cumsum_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_fill_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_logsumexp_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_mean_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_scatter_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_scatter_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_select_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_var_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_var_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_meshgrid_variadic_tensors_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mm_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_movedim_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_multinomial_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mvlgamma_mvlgamma_p_3_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nanmedian_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nansum_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nansum_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_narrow_copy_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_narrow_copy_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_ne_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_new_empty_strided_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_new_zeros_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_adaptive_avg_pool2d_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_alpha_dropout_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_avg_pool3d_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_binary_cross_entropy_with_logits_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_conv1d_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_conv2d_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_conv_transpose3d_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_embedding_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_feature_alpha_dropout_without_train_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_fractional_max_pool2d_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_hardsigmoid_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_hardswish_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_instance_norm_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_logsigmoid_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_logsigmoid_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_max_unpool2d_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_max_unpool3d_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_multi_head_attention_forward_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pad_replicate_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pad_replicate_negative_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pad_replicate_negative_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pixel_shuffle_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pixel_shuffle_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_relu6_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_smooth_l1_loss_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_softsign_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_softsign_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_upsample_nearest_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nonzero_static_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_norm_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_normal_in_place_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_ones_like_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_ones_like_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_permute_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_permute_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_pinverse_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_polygamma_polygamma_n_0_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_polygamma_polygamma_n_0_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_positive_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_randint_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_randn_like_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_renorm_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_repeat_interleave_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_resize__cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_roll_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_rot90_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_round_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scalar_tensor_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scatter_add_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scatter_reduce_amax_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scatter_reduce_amax_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scatter_reduce_prod_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_short_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sign_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_signbit_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_signbit_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sinh_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_slice_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_softmax_with_dtype_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_softmax_with_dtype_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_bessel_y1_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_chebyshev_polynomial_w_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_hermite_polynomial_h_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_hermite_polynomial_h_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_hermite_polynomial_he_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_i1_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_i1_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_i1e_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_log_ndtr_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_modified_bessel_i1_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_ndtr_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_scaled_modified_bessel_k1_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_spherical_bessel_j0_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_xlog1py_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_split_with_sizes_copy_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_split_with_sizes_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sqrt_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_squeeze_multiple_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_stack_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_std_mean_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_std_unbiased_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sum_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sum_to_size_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_take_along_dim_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_tan_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_tan_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_tensor_split_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_trace_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_trace_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_trapezoid_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_tril_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_triu_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_triu_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unflatten_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unfold_copy_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unique_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unsafe_chunk_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unsafe_split_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unsqueeze_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_vdot_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_view_as_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_view_copy_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_vstack_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_where_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_xlogy_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_zero__cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_zeros_like_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick__chunk_cat_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick__unsafe_masked_index_put_accumulate_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_abs_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_add_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_add_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_addmv_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_addmv_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_aminmax_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_as_strided_copy_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_atan2_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_atan_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_baddbmm_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_bitwise_left_shift_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_bitwise_not_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_bitwise_right_shift_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_bitwise_xor_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_cat_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_cat_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_conj_physical_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_constant_pad_nd_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_index_fill_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_cosh_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_cosh_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_count_nonzero_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_cumprod_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_diag_embed_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_diagonal_copy_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_diagonal_scatter_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_div_no_rounding_mode_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_eq_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_erf_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_erfc_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_expm1_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_fft_fft2_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_fft_fft_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_fft_ihfft_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_fft_ihfftn_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_fft_ihfftn_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_fft_irfft_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_fft_irfftn_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_fft_rfft_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_fmin_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_i0_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_i0_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_i0_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_index_add_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_index_add_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_index_select_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_index_select_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_isneginf_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_linalg_diagonal_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_linalg_vector_norm_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_linalg_vector_norm_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_log10_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_log_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_log_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_logaddexp2_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_logical_not_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_logit_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_logspace_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_maximum_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_mul_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_narrow_copy_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_ne_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_new_empty_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_new_empty_strided_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_new_full_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_new_zeros_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_new_zeros_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_leaky_relu_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_pad_constant_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_relu_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_relu_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_silu_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_normal_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_normal_in_place_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_ones_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_ones_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_permute_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_pow_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_prod_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_rad2deg_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_reciprocal_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_remainder_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_rot90_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_sigmoid_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_sigmoid_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_sigmoid_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_signbit_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_sinc_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_slice_scatter_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_special_i1e_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_split_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_split_list_args_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_split_list_args_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_split_with_sizes_copy_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_squeeze_multiple_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_std_mean_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_sum_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_t_copy_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_take_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_tan_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_tanh_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_tril_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_quick_triu_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_triu_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_triu_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_unbind_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_unfold_copy_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_unsqueeze_copy_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_vdot_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_view_copy_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_view_copy_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_view_copy_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_zeros_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_zeros_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_rnn_decomp_module_nn_GRU_eval_mode_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_rnn_decomp_module_nn_GRU_train_mode_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_rrelu_with_noise_cuda 2024-08-06T21:11:09.8915401Z 2024-08-06T21:11:13.5438062Z Running test_model_exports_to_core_aten 1/1 ... [2024-08-06 21:11:13.543291] 2024-08-06T21:11:13.5441906Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_model_exports_to_core_aten.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:11:13.543798] 2024-08-06T21:11:17.7661567Z 2024-08-06T21:11:17.7663168Z test_model_exports_to_core_aten 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_model_exports_to_core_aten_1.1_17753ab4128f7cce_.log 2024-08-06T21:11:17.7664606Z Running 1 items in this shard: test/test_model_exports_to_core_aten.py::TestQuantizePT2EModels::test_vit_aten_export 2024-08-06T21:11:17.7665208Z 2024-08-06T21:11:21.4768903Z Running test_transformers 1/1 ... [2024-08-06 21:11:21.476341] 2024-08-06T21:11:21.4773325Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_transformers.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:11:21.476897] 2024-08-06T21:14:57.1889976Z 2024-08-06T21:14:57.1890903Z test_decomp 24/24 was successful, full logs can be found in artifacts with path test/test-reports/test_decomp_24.24_c4055e2d41561e4a_.log 2024-08-06T21:14:57.2026309Z Running 361 items in this shard: test/test_decomp.py::TestDecompCUDA::test_comprehensive___getitem___cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive___getitem___cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive___radd___cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive___radd___cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive___rdiv___cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive___rmatmul___cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive___rmatmul___cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive___rmod___cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive___rmul___cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive__chunk_cat_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive__unsafe_masked_index_put_accumulate_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive__unsafe_masked_index_put_accumulate_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_abs_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_allclose_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_amin_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_aminmax_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_as_strided_copy_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_as_strided_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_asin_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_asinh_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_atanh_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_atleast_1d_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_baddbmm_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_bfloat16_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_bitwise_left_shift_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_block_diag_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_bool_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_bucketize_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cat_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cat_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cfloat_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_char_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_column_stack_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_conj_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_constant_pad_nd_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_copysign_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_corrcoef_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cos_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cosh_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cov_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cov_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cov_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cummax_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_cumprod_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_deg2rad_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_deg2rad_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diag_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diagflat_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diagonal_copy_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diagonal_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diagonal_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_diagonal_scatter_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_double_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_empty_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_empty_permuted_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_equal_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_equal_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_exp2_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_expm1_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_fft2_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_hfft_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_ifft2_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_ifft_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_ihfftn_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_ihfftn_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_irfft2_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fft_rfft_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fill_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_flip_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_flip_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fliplr_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fliplr_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_float_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_floor_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fmax_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_fmod_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_geometric_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_gt_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_histc_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_hsplit_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_hstack_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_hstack_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_index_add_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_index_copy_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_index_reduce_prod_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_index_select_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_int_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_int_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_isnan_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_isneginf_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_jiterator_2inputs_2outputs_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_jiterator_2inputs_2outputs_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_jiterator_4inputs_with_extra_args_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_jiterator_4inputs_with_extra_args_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_jiterator_unary_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_lcm_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_lcm_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_lgamma_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_cholesky_ex_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_diagonal_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_pinv_singular_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linalg_solve_ex_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linspace_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linspace_tensor_overload_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_linspace_tensor_overload_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_log10_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_log10_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_log1p_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_log2_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logical_and_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logical_and_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logical_and_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logical_or_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logspace_tensor_overload_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logspace_tensor_overload_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_logsumexp_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_lu_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_lu_solve_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_amin_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_amin_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_argmax_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_argmax_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_log_softmax_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_log_softmax_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_normalize_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_normalize_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_scatter_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_select_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_masked_select_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_matrix_exp_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_min_reduction_with_dim_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_minimum_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mm_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mode_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_msort_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mvlgamma_mvlgamma_p_3_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mvlgamma_mvlgamma_p_3_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_mvlgamma_mvlgamma_p_5_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nansum_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_narrow_copy_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_native_batch_norm_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_neg_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_neg_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_new_full_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nextafter_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_adaptive_max_pool3d_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_batch_norm_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_celu_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_channel_shuffle_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_conv_transpose2d_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_cosine_similarity_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_dropout2d_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_dropout2d_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_embedding_bag_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_gelu_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_group_norm_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_huber_loss_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_interpolate_area_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_interpolate_nearest-exact_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_linear_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_max_pool2d_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_multi_head_attention_forward_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pad_reflect_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pad_replicate_negative_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pixel_shuffle_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pixel_shuffle_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pixel_unshuffle_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_pixel_unshuffle_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_poisson_nll_loss_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_prelu_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_scaled_dot_product_attention_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_nn_functional_tanhshrink_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_ones_like_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_polygamma_polygamma_n_1_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_polygamma_polygamma_n_2_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_polygamma_polygamma_n_2_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_polygamma_polygamma_n_3_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_pow_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_rad2deg_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_randint_like_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_reciprocal_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_repeat_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_repeat_interleave_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_reshape_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_resize_as__cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_resize_as__cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_resolve_conj_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_roll_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_rot90_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_round_decimals_0_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_rsqrt_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scatter_add_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scatter_reduce_mean_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_scatter_reduce_sum_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_searchsorted_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sign_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sign_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_signal_windows_cosine_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_signal_windows_hamming_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_sinh_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_slice_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_slice_scatter_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_airy_ai_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_entr_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_erfcx_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_hermite_polynomial_h_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_laguerre_polynomial_l_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_legendre_polynomial_p_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_modified_bessel_i1_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_modified_bessel_k0_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_modified_bessel_k1_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_scaled_modified_bessel_k1_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_special_xlog1py_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_comprehensive_split_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_square_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_squeeze_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_std_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_std_mean_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_std_unbiased_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_t_copy_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_tan_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_topk_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_trapezoid_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_triangular_solve_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_tril_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_triu_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_true_divide_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_trunc_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unbind_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unbind_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unflatten_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unfold_copy_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unfold_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unique_consecutive_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unique_consecutive_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unique_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unique_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_unravel_index_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_comprehensive_var_mean_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_var_mean_unbiased_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_comprehensive_vdot_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_comprehensive_view_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_comprehensive_vstack_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_comprehensive_vstack_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick__unsafe_masked_index_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_abs_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_quick_acos_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_acosh_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_addr_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_addr_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_alias_copy_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_all_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_any_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_as_strided_scatter_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_bitwise_and_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_bucketize_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_ceil_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_clamp_min_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_clone_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_addcmul_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_norm_inf_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_sgn_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_core_backward_std_mean_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_cos_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_cosh_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_count_nonzero_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_cumprod_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_diagonal_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_div_no_rounding_mode_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_empty_like_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_quick_empty_like_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_empty_strided_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_eq_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_erfinv_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_fft_fft_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_fft_hfft2_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_fft_ihfft2_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_fft_ihfft_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_fft_ihfft_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_fft_rfft2_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_fft_rfftn_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_flip_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_flip_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_floor_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_fmod_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_full_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_gcd_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_ge_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_geometric_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_grid_sampler_2d_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_gt_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_index_copy_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_index_fill_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_isinf_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_linalg_cross_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_linalg_cross_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_linalg_diagonal_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_linspace_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_log10_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_logaddexp2_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_logical_not_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_logical_xor_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_logical_xor_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_logit_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_logspace_tensor_overload_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_logspace_tensor_overload_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_logspace_tensor_overload_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_logsumexp_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_meshgrid_list_of_tensors_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_meshgrid_list_of_tensors_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_meshgrid_variadic_tensors_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_meshgrid_variadic_tensors_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_meshgrid_variadic_tensors_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_mul_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_mvlgamma_mvlgamma_p_1_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_mvlgamma_mvlgamma_p_3_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_nan_to_num_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_nansum_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_ne_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_ne_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_new_empty_strided_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_new_full_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_new_ones_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_hardtanh_cuda_int8, test/test_decomp.py::TestDecompCUDA::test_quick_nn_functional_silu_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_normal_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_ones_like_cuda_int32, test/test_decomp.py::TestDecompCUDA::test_quick_prod_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_prod_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_randn_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_repeat_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_repeat_cuda_int64, test/test_decomp.py::TestDecompCUDA::test_quick_roll_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_rsub_cuda_complex128, test/test_decomp.py::TestDecompCUDA::test_quick_rsub_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_select_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_sign_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_split_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_quick_split_with_sizes_copy_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_sqrt_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_std_mean_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_sub_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_sub_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_t_copy_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_t_copy_cuda_float16, test/test_decomp.py::TestDecompCUDA::test_quick_take_cuda_int16, test/test_decomp.py::TestDecompCUDA::test_quick_tanh_cuda_complex32, test/test_decomp.py::TestDecompCUDA::test_quick_trace_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_transpose_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_quick_unfold_copy_cuda_complex64, test/test_decomp.py::TestDecompCUDA::test_quick_unfold_cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_unfold_cuda_bool, test/test_decomp.py::TestDecompCUDA::test_quick_unsqueeze_copy_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_var_mean_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_view_copy_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_view_cuda_float64, test/test_decomp.py::TestDecompCUDA::test_quick_view_cuda_uint8, test/test_decomp.py::TestDecompCUDA::test_quick_zero__cuda_bfloat16, test/test_decomp.py::TestDecompCUDA::test_quick_zeros_like_cuda_float32, test/test_decomp.py::TestDecompCUDA::test_rnn_decomp_module_nn_RNN_eval_mode_cuda_float64, test/test_decomp.py::DecompOneOffTestsCUDA::test_contiguous_softmax_cuda 2024-08-06T21:14:57.2156672Z 2024-08-06T21:15:00.8407957Z Running functorch/test_ops 8/9 ... [2024-08-06 21:15:00.840230] 2024-08-06T21:15:00.8412587Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'functorch/test_ops.py', '-m', 'not serial', '--shard-id=8', '--num-shards=9', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:15:00.840736] 2024-08-06T21:15:36.4685953Z 2024-08-06T21:15:36.4686987Z PRINTING LOG FILE of test_transformers 1/1 (test/test-reports/test_transformers_1.1_95af2e2cb07b8232_.log) 2024-08-06T21:15:36.4832814Z Test results will be stored in test-reports/python-pytest/test_transformers/test_transformers-cd06bb4e6af626fa.xml 2024-08-06T21:15:36.4834216Z ============================= test session starts ============================== 2024-08-06T21:15:36.4835255Z platform linux -- Python 3.10.14, pytest-7.3.2, pluggy-1.5.0 -- /opt/conda/envs/py_3.10/bin/python 2024-08-06T21:15:36.4835913Z cachedir: .pytest_cache 2024-08-06T21:15:36.4836867Z hypothesis profile 'pytorch_ci' -> database=None, max_examples=50, derandomize=True, suppress_health_check=[HealthCheck.too_slow] 2024-08-06T21:15:36.4837665Z rootdir: /var/lib/jenkins/workspace 2024-08-06T21:15:36.4838067Z configfile: pytest.ini 2024-08-06T21:15:36.4838694Z plugins: hypothesis-5.35.1, cpp-2.3.0, flakefinder-1.1.0, rerunfailures-14.0, xdist-3.3.1, xdoctest-1.1.0 2024-08-06T21:15:36.4840207Z collecting ... collected 45344 items 2024-08-06T21:15:36.4841073Z stepcurrent: Cannot find last run test, not skipping 2024-08-06T21:15:40.6222781Z Running 45344 items in this shard: test/test_transformers.py::TestSDPAPrivateUse1Only::test_fused_sdp_choice_privateuseone, test/test_transformers.py::TestSDPAPrivateUse1Only::test_scaled_dot_product_fused_attention_overrideable, test/test_transformers.py::TestSDPAPrivateUse1Only::test_scaled_dot_product_fused_attention_overrideable_backward, test/test_transformers.py::TestTransformersCUDA::test_bias_is_none_cuda, test/test_transformers.py::TestTransformersCUDA::test_decoder_only_layer_cuda, test/test_transformers.py::TestTransformersCUDA::test_decoder_padding_and_src_mask_bool_cuda, test/test_transformers.py::TestTransformersCUDA::test_disable_fastpath_cuda, test/test_transformers.py::TestTransformersCUDA::test_encoder_is_causal_cuda, test/test_transformers.py::TestTransformersCUDA::test_encoder_padding_and_src_mask_bool_cuda, test/test_transformers.py::TestTransformersCUDA::test_is_causal_gpu_cuda, test/test_transformers.py::TestTransformersCUDA::test_kpm_mask_trailing_column_with_nested_tensor_cuda, test/test_transformers.py::TestTransformersCUDA::test_mask_check_fastpath_cuda, test/test_transformers.py::TestTransformersCUDA::test_mha_native_args_nb_heads_1_bias_False_cuda, test/test_transformers.py::TestTransformersCUDA::test_mha_native_args_nb_heads_1_bias_True_cuda, test/test_transformers.py::TestTransformersCUDA::test_mha_native_args_nb_heads_8_bias_False_cuda, test/test_transformers.py::TestTransformersCUDA::test_mha_native_args_nb_heads_8_bias_True_cuda, test/test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim2_key_padding_mask_dim1_bool_cuda, test/test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim2_key_padding_mask_dim1_float32_cuda, test/test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim2_key_padding_mask_dim_2_bool_cuda, test/test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim2_key_padding_mask_dim_2_float32_cuda, test/test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_2_key_padding_mask_dim1_bool_cuda, test/test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_2_key_padding_mask_dim1_float32_cuda, test/test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_2_key_padding_mask_dim_2_bool_cuda, test/test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_2_key_padding_mask_dim_2_float32_cuda, test/test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_3_key_padding_mask_dim1_bool_cuda, test/test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_3_key_padding_mask_dim1_float32_cuda, test/test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_3_key_padding_mask_dim_2_bool_cuda, test/test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_3_key_padding_mask_dim_2_float32_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_2D_attn_mask_dropout_p_0_0_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_2D_attn_mask_dropout_p_0_2_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_2D_attn_mask_dropout_p_0_5_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_2D_causal_attn_mask_dropout_p_0_0_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_2D_causal_attn_mask_dropout_p_0_2_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_2D_causal_attn_mask_dropout_p_0_5_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_3D_attn_mask_dropout_p_0_0_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_3D_attn_mask_dropout_p_0_2_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_3D_attn_mask_dropout_p_0_5_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_3D_causal_attn_mask_dropout_p_0_0_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_3D_causal_attn_mask_dropout_p_0_2_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_3D_causal_attn_mask_dropout_p_0_5_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_no_attn_mask_dropout_p_0_0_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_no_attn_mask_dropout_p_0_2_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_no_attn_mask_dropout_p_0_5_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_2D_attn_mask_dropout_p_0_0_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_2D_attn_mask_dropout_p_0_2_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_2D_attn_mask_dropout_p_0_5_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_2D_causal_attn_mask_dropout_p_0_0_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_2D_causal_attn_mask_dropout_p_0_2_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_2D_causal_attn_mask_dropout_p_0_5_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_4D_attn_mask_dropout_p_0_0_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_4D_attn_mask_dropout_p_0_2_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_4D_attn_mask_dropout_p_0_5_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_4D_causal_attn_mask_dropout_p_0_0_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_4D_causal_attn_mask_dropout_p_0_2_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_4D_causal_attn_mask_dropout_p_0_5_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_no_attn_mask_dropout_p_0_0_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_no_attn_mask_dropout_p_0_2_cuda, test/test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_no_attn_mask_dropout_p_0_5_cuda, test/test_transformers.py::TestTransformersCUDA::test_script_encoder_subclass_cuda, test/test_transformers.py::TestTransformersCUDA::test_script_mha_in_proj_weight_none_cuda, test/test_transformers.py::TestTransformersCUDA::test_self_attn_TxT_attn_mask_cuda, test/test_transformers.py::TestTransformersCUDA::test_train_with_is_causal_cuda, test/test_transformers.py::TestTransformersCUDA::test_train_with_pad_and_catch_error_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformer_bias_is_none_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_False_training_False_enable_nested_tensor_False_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_False_training_False_enable_nested_tensor_True_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_False_training_True_enable_nested_tensor_False_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_False_training_True_enable_nested_tensor_True_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_True_training_False_enable_nested_tensor_False_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_True_training_False_enable_nested_tensor_True_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_True_training_True_enable_nested_tensor_False_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_True_training_True_enable_nested_tensor_True_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_False_use_autocast_False_d_model_12_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_False_use_autocast_False_d_model_256_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_False_use_autocast_True_d_model_12_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_False_use_autocast_True_d_model_256_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_True_use_autocast_False_d_model_12_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_True_use_autocast_False_d_model_256_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_True_use_autocast_True_d_model_12_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_True_use_autocast_True_d_model_256_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_square_input_with_no_grad_False_training_False_enable_nested_tensor_False_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_square_input_with_no_grad_False_training_True_enable_nested_tensor_False_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_square_input_with_no_grad_True_training_False_enable_nested_tensor_False_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoder_square_input_with_no_grad_True_training_True_enable_nested_tensor_False_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoderlayer_no_fastpath_with_hooks_nhead_3_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoderlayer_no_fastpath_with_hooks_nhead_4_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoderlayer_src_mask_nhead_1_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoderlayer_src_mask_nhead_4_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoderlayer_src_mask_nhead_8_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoderlayer_subclass_cuda, test/test_transformers.py::TestTransformersCUDA::test_transformerencoderlayer_subclass_model_cuda, test/test_transformers.py::TestTransformersCUDA::test_with_nested_tensor_input_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_dispatch_fails_no_backend_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_flash_atteention_large_bf16_nan_values_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_flash_attention_fail_with_non_square_causal_attention_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_flash_autocast_fp32_bfloat16_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_flash_autocast_fp32_float16_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_flash_backward_failure_sm86plus_head_dim_193_dropout_p_0_0_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_flash_backward_failure_sm86plus_head_dim_193_dropout_p_0_2_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_flash_backward_failure_sm86plus_head_dim_204_dropout_p_0_0_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_flash_backward_failure_sm86plus_head_dim_204_dropout_p_0_2_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_flash_backward_failure_sm86plus_head_dim_256_dropout_p_0_0_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_flash_backward_failure_sm86plus_head_dim_256_dropout_p_0_2_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_flash_fail_fp32_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_fused_kernels_nested_broadcasting_error_cases_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_fused_kernels_nested_broadcasting_requires_grad_failure_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_fused_kernels_seq_len_0_inputs_fused_kernel0_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_fused_inputs_attn_mask_present_kernel0_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_fused_inputs_broadcast_kernel0_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_fused_inputs_dim_3_kernel0_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_fused_inputs_head_dim_kernel0_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_fused_inputs_invalid_dtype_kernel0_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_1_dimensional_inputs_kernel0_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_1_dimensional_inputs_kernel1_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_1_dimensional_inputs_kernel2_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_different_datatypes_kernel0_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_different_datatypes_kernel1_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_different_datatypes_kernel2_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_different_devices_kernel0_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_different_devices_kernel1_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_different_devices_kernel2_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_last_dim_stride_kernel0_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_sequence_lengths_kernel0_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_mem_efficient_fail_bfloat16_less_than_sm80_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_nested_fails_on_padding_head_dim_cuda, test/test_transformers.py::TestSDPAFailureModesCUDA::test_unaligned_tensors_cuda, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_0_bfloat16_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_0_float16_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_0_float32_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_0_float64_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_7_bfloat16_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_7_float16_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_7_float32_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_7_float64_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_0_bfloat16_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_0_float16_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_0_float32_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_0_float64_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_7_bfloat16_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_7_float16_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_7_float32_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_7_float64_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_attention_math_with_negative_scale_kernel0_cuda, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_bfloat16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_float16, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_float32, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_float64, test/test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_with_inf_cuda, test/test_transformers.py::TestSDPACUDA::test_sdp_math_gradcheck_contiguous_inputs_False_cuda, test/test_transformers.py::TestSDPACUDA::test_sdp_math_gradcheck_contiguous_inputs_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_cudnn_attention_different_dk_dv_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_cudnn_attention_fail_d128_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_32_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_32_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_32_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_32_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_64_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_64_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_64_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_64_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_32_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_32_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_32_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_32_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_64_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_64_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_64_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_64_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_32_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_32_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_32_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_32_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_64_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_64_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_64_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_64_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_32_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_32_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_32_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_32_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_64_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_64_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_64_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_64_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_32_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_32_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_32_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_32_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_64_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_64_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_64_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_64_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_32_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_32_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_32_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_32_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_64_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_64_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_64_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_64_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_backwards_throws_determinism_warning_fused_kernel0_warn_only_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_backwards_throws_determinism_warning_fused_kernel0_warn_only_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_query_dense_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_seq_len_1_inputs_fused_kernel0_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_sdp_choice_type_dense_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_sdp_choice_type_nested_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_attention_long_sequence_mask_float16_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_attention_long_sequence_mask_float32_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_attention_non_contig_mask_bug_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_attention_non_contiguous_mask_float16_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_attention_non_contiguous_mask_float32_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_attention_pad_mask_float16_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_attention_pad_mask_float32_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_backwards_determinism_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_mask_variants_mask_dim_1_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_mask_variants_mask_dim_2_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_mask_variants_mask_dim_3_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_mask_variants_mask_dim_4_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_scaled_dot_product_attention_fused_kernels_packed_accuracy_type_dense_fused_kernel0_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_scaled_dot_product_attention_fused_kernels_packed_accuracy_type_nested_fused_kernel0_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_scaled_dot_product_attention_fused_kernels_packed_type_dense_is_contiguous_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_scaled_dot_product_attention_fused_kernels_packed_type_dense_is_contiguous_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_scaled_dot_product_attention_fused_kernels_packed_type_nested_is_contiguous_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_scaled_dot_product_attention_fused_kernels_packed_type_nested_is_contiguous_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_choice_with_determinism_warn_only_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_choice_with_determinism_warn_only_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_False_is_causal_False_bfloat16_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_False_is_causal_False_float16_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_False_is_causal_True_bfloat16_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_False_is_causal_True_float16_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_True_is_causal_False_bfloat16_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_True_is_causal_False_float16_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_True_is_causal_True_bfloat16_cuda_bfloat16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_True_is_causal_True_float16_cuda_float16, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_mem_efficient_grad_against_math_contiguous_inputs_False_is_causal_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_mem_efficient_grad_against_math_contiguous_inputs_False_is_causal_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_mem_efficient_grad_against_math_contiguous_inputs_True_is_causal_False_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_mem_efficient_grad_against_math_contiguous_inputs_True_is_causal_True_cuda, test/test_transformers.py::TestSDPACudaOnlyCUDA::test_singelton_head_dim_stride_ne_1_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_LOWER_RIGHT_shape0_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_LOWER_RIGHT_shape1_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_LOWER_RIGHT_shape2_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_LOWER_RIGHT_shape3_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_UPPER_LEFT_shape0_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_UPPER_LEFT_shape1_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_UPPER_LEFT_shape2_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_UPPER_LEFT_shape3_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_LOWER_RIGHT_shape0_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_LOWER_RIGHT_shape1_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_LOWER_RIGHT_shape2_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_LOWER_RIGHT_shape3_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_UPPER_LEFT_shape0_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_UPPER_LEFT_shape1_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_UPPER_LEFT_shape2_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_UPPER_LEFT_shape3_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_is_causal_and_mask_fails_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_is_causal_equals_upper_left_shape0_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_is_causal_equals_upper_left_shape1_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_is_causal_equals_upper_left_shape2_cuda, test/test_transformers.py::TestAttnBiasCUDA::test_is_causal_equals_upper_left_shape3_cuda 2024-08-06T21:15:44.6654256Z 2024-08-06T21:15:44.6673112Z test_transformers.py::TestSDPAPrivateUse1Only::test_fused_sdp_choice_privateuseone [1/2] c++ -MMD -MF open_registration_extension.o.d -DTORCH_EXTENSION_NAME=custom_device_extension -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE=\"_gcc\" -DPYBIND11_STDLIB=\"_libstdcpp\" -DPYBIND11_BUILD_ABI=\"_cxxabi1013\" -I/var/lib/jenkins/workspace/test/cpp_extensions -isystem /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/include -isystem /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/include/torch/csrc/api/include -isystem /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/include/TH -isystem /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/include/THC -isystem /opt/conda/envs/py_3.10/include/python3.10 -D_GLIBCXX_USE_CXX11_ABI=1 -fPIC -std=c++17 -g -c /var/lib/jenkins/workspace/test/cpp_extensions/open_registration_extension.cpp -o open_registration_extension.o 2024-08-06T21:15:44.6678915Z [2/2] c++ open_registration_extension.o -shared -L/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib -lc10 -ltorch_cpu -ltorch -ltorch_python -o custom_device_extension.so 2024-08-06T21:15:44.6680557Z PASSED [0.0127s] [ 0%] 2024-08-06T21:15:44.6681736Z test_transformers.py::TestSDPAPrivateUse1Only::test_scaled_dot_product_fused_attention_overrideable PASSED [0.0114s] [ 0%] 2024-08-06T21:15:44.6683409Z test_transformers.py::TestSDPAPrivateUse1Only::test_scaled_dot_product_fused_attention_overrideable_backward PASSED [0.0219s] [ 0%] 2024-08-06T21:15:44.6684916Z test_transformers.py::TestTransformersCUDA::test_bias_is_none_cuda PASSED [0.0053s] [ 0%] 2024-08-06T21:15:44.6686327Z test_transformers.py::TestTransformersCUDA::test_decoder_only_layer_cuda SKIPPED [0.0003s] (Fairseq not found) [ 0%] 2024-08-06T21:15:44.6688030Z test_transformers.py::TestTransformersCUDA::test_decoder_padding_and_src_mask_bool_cuda SKIPPED [0.0002s] (not supported on pre-3.11 Python) [ 0%] 2024-08-06T21:15:44.6689656Z test_transformers.py::TestTransformersCUDA::test_disable_fastpath_cuda PASSED [0.2704s] [ 0%] 2024-08-06T21:15:44.6690916Z test_transformers.py::TestTransformersCUDA::test_encoder_is_causal_cuda PASSED [0.0058s] [ 0%] 2024-08-06T21:15:44.6692488Z test_transformers.py::TestTransformersCUDA::test_encoder_padding_and_src_mask_bool_cuda SKIPPED [0.0002s] (not supported on pre-3.11 Python) [ 0%] 2024-08-06T21:15:44.6694466Z test_transformers.py::TestTransformersCUDA::test_is_causal_gpu_cuda SKIPPED [0.0002s] (Platform does not supposrt fused SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.6696213Z test_transformers.py::TestTransformersCUDA::test_kpm_mask_trailing_column_with_nested_tensor_cuda PASSED [0.0404s] [ 0%] 2024-08-06T21:15:44.6697611Z test_transformers.py::TestTransformersCUDA::test_mask_check_fastpath_cuda PASSED [0.0151s] [ 0%] 2024-08-06T21:15:44.6699143Z test_transformers.py::TestTransformersCUDA::test_mha_native_args_nb_heads_1_bias_False_cuda PASSED [0.0046s] [ 0%] 2024-08-06T21:15:44.6700879Z test_transformers.py::TestTransformersCUDA::test_mha_native_args_nb_heads_1_bias_True_cuda PASSED [0.0046s] [ 0%] 2024-08-06T21:15:44.6702406Z test_transformers.py::TestTransformersCUDA::test_mha_native_args_nb_heads_8_bias_False_cuda PASSED [0.0039s] [ 0%] 2024-08-06T21:15:44.6703870Z test_transformers.py::TestTransformersCUDA::test_mha_native_args_nb_heads_8_bias_True_cuda PASSED [0.0224s] [ 0%] 2024-08-06T21:15:44.6706534Z test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim2_key_padding_mask_dim1_bool_cuda PASSED [0.0384s] [ 0%] 2024-08-06T21:15:44.6709019Z test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim2_key_padding_mask_dim1_float32_cuda PASSED [0.0035s] [ 0%] 2024-08-06T21:15:44.6711206Z test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim2_key_padding_mask_dim_2_bool_cuda PASSED [0.0092s] [ 0%] 2024-08-06T21:15:44.6713105Z test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim2_key_padding_mask_dim_2_float32_cuda PASSED [0.0035s] [ 0%] 2024-08-06T21:15:44.6715117Z test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_2_key_padding_mask_dim1_bool_cuda PASSED [0.0035s] [ 0%] 2024-08-06T21:15:44.6717156Z test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_2_key_padding_mask_dim1_float32_cuda PASSED [0.0033s] [ 0%] 2024-08-06T21:15:44.6719124Z test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_2_key_padding_mask_dim_2_bool_cuda PASSED [0.0035s] [ 0%] 2024-08-06T21:15:44.6721081Z test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_2_key_padding_mask_dim_2_float32_cuda PASSED [0.0035s] [ 0%] 2024-08-06T21:15:44.6723098Z test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_3_key_padding_mask_dim1_bool_cuda PASSED [0.0033s] [ 0%] 2024-08-06T21:15:44.6725054Z test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_3_key_padding_mask_dim1_float32_cuda PASSED [0.0033s] [ 0%] 2024-08-06T21:15:44.6727084Z test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_3_key_padding_mask_dim_2_bool_cuda PASSED [0.0035s] [ 0%] 2024-08-06T21:15:44.6729002Z test_transformers.py::TestTransformersCUDA::test_multiheadattention_fastpath_attn_mask_attn_mask_dim_3_key_padding_mask_dim_2_float32_cuda PASSED [0.0035s] [ 0%] 2024-08-06T21:15:44.6730890Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_2D_attn_mask_dropout_p_0_0_cuda PASSED [0.0083s] [ 0%] 2024-08-06T21:15:44.6732752Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_2D_attn_mask_dropout_p_0_2_cuda PASSED [0.0040s] [ 0%] 2024-08-06T21:15:44.6734536Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_2D_attn_mask_dropout_p_0_5_cuda PASSED [0.0039s] [ 0%] 2024-08-06T21:15:44.6736372Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_2D_causal_attn_mask_dropout_p_0_0_cuda PASSED [0.0091s] [ 0%] 2024-08-06T21:15:44.6738309Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_2D_causal_attn_mask_dropout_p_0_2_cuda PASSED [0.0044s] [ 0%] 2024-08-06T21:15:44.6740166Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_2D_causal_attn_mask_dropout_p_0_5_cuda PASSED [0.0044s] [ 0%] 2024-08-06T21:15:44.6741968Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_3D_attn_mask_dropout_p_0_0_cuda PASSED [0.0038s] [ 0%] 2024-08-06T21:15:44.6743837Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_3D_attn_mask_dropout_p_0_2_cuda PASSED [0.0039s] [ 0%] 2024-08-06T21:15:44.6745569Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_3D_attn_mask_dropout_p_0_5_cuda PASSED [0.0039s] [ 0%] 2024-08-06T21:15:44.6747504Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_3D_causal_attn_mask_dropout_p_0_0_cuda PASSED [0.0043s] [ 0%] 2024-08-06T21:15:44.6749563Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_3D_causal_attn_mask_dropout_p_0_2_cuda PASSED [0.0044s] [ 0%] 2024-08-06T21:15:44.6751457Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_3D_causal_attn_mask_dropout_p_0_5_cuda PASSED [0.0045s] [ 0%] 2024-08-06T21:15:44.6754892Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_no_attn_mask_dropout_p_0_0_cuda SKIPPED [0.0007s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131086 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 0%] 2024-08-06T21:15:44.6759559Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_no_attn_mask_dropout_p_0_2_cuda SKIPPED [0.0007s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131146 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 0%] 2024-08-06T21:15:44.6764152Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_3D_input_dim_no_attn_mask_dropout_p_0_5_cuda SKIPPED [0.0008s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131123 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 0%] 2024-08-06T21:15:44.6767250Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_2D_attn_mask_dropout_p_0_0_cuda PASSED [0.0036s] [ 0%] 2024-08-06T21:15:44.6769010Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_2D_attn_mask_dropout_p_0_2_cuda PASSED [0.0040s] [ 0%] 2024-08-06T21:15:44.6770789Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_2D_attn_mask_dropout_p_0_5_cuda PASSED [0.0036s] [ 0%] 2024-08-06T21:15:44.6772852Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_2D_causal_attn_mask_dropout_p_0_0_cuda PASSED [0.0043s] [ 0%] 2024-08-06T21:15:44.6774719Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_2D_causal_attn_mask_dropout_p_0_2_cuda PASSED [0.0041s] [ 0%] 2024-08-06T21:15:44.6776534Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_2D_causal_attn_mask_dropout_p_0_5_cuda PASSED [0.0041s] [ 0%] 2024-08-06T21:15:44.6778428Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_4D_attn_mask_dropout_p_0_0_cuda PASSED [0.0035s] [ 0%] 2024-08-06T21:15:44.6780205Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_4D_attn_mask_dropout_p_0_2_cuda PASSED [0.0037s] [ 0%] 2024-08-06T21:15:44.6781956Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_4D_attn_mask_dropout_p_0_5_cuda PASSED [0.0036s] [ 0%] 2024-08-06T21:15:44.6783959Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_4D_causal_attn_mask_dropout_p_0_0_cuda PASSED [0.0040s] [ 0%] 2024-08-06T21:15:44.6785804Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_4D_causal_attn_mask_dropout_p_0_2_cuda PASSED [0.0042s] [ 0%] 2024-08-06T21:15:44.6787753Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_4D_causal_attn_mask_dropout_p_0_5_cuda PASSED [0.0041s] [ 0%] 2024-08-06T21:15:44.6791288Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_no_attn_mask_dropout_p_0_0_cuda SKIPPED [0.0007s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/129853 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 0%] 2024-08-06T21:15:44.6795915Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_no_attn_mask_dropout_p_0_2_cuda SKIPPED [0.0007s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131107 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 0%] 2024-08-06T21:15:44.6800529Z test_transformers.py::TestTransformersCUDA::test_scaled_dot_product_attention_4D_input_dim_no_attn_mask_dropout_p_0_5_cuda SKIPPED [0.0007s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131179 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 0%] 2024-08-06T21:15:44.6803385Z test_transformers.py::TestTransformersCUDA::test_script_encoder_subclass_cuda PASSED [0.4001s] [ 0%] 2024-08-06T21:15:44.6805300Z test_transformers.py::TestTransformersCUDA::test_script_mha_in_proj_weight_none_cuda PASSED [0.0343s] [ 0%] 2024-08-06T21:15:44.6807074Z test_transformers.py::TestTransformersCUDA::test_self_attn_TxT_attn_mask_cuda SKIPPED [0.0003s] (4D mask not supported yet - activate when 4D mask supported) [ 0%] 2024-08-06T21:15:44.6808723Z test_transformers.py::TestTransformersCUDA::test_train_with_is_causal_cuda PASSED [0.0495s] [ 0%] 2024-08-06T21:15:44.6810428Z test_transformers.py::TestTransformersCUDA::test_train_with_pad_and_catch_error_cuda SKIPPED [0.0014s] (test is slow; run with PYTORCH_TEST_WITH_SLOW to enable test) [ 0%] 2024-08-06T21:15:44.6812060Z test_transformers.py::TestTransformersCUDA::test_transformer_bias_is_none_cuda PASSED [0.0449s] [ 0%] 2024-08-06T21:15:44.6813809Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_False_training_False_enable_nested_tensor_False_cuda PASSED [0.0469s] [ 0%] 2024-08-06T21:15:44.6815766Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_False_training_False_enable_nested_tensor_True_cuda PASSED [0.0464s] [ 0%] 2024-08-06T21:15:44.6817791Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_False_training_True_enable_nested_tensor_False_cuda PASSED [0.0494s] [ 0%] 2024-08-06T21:15:44.6819667Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_False_training_True_enable_nested_tensor_True_cuda PASSED [0.0500s] [ 0%] 2024-08-06T21:15:44.6821615Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_True_training_False_enable_nested_tensor_False_cuda PASSED [0.0398s] [ 0%] 2024-08-06T21:15:44.6823591Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_True_training_False_enable_nested_tensor_True_cuda PASSED [0.0664s] [ 0%] 2024-08-06T21:15:44.6825520Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_True_training_True_enable_nested_tensor_False_cuda PASSED [0.0512s] [ 0%] 2024-08-06T21:15:44.6827534Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_batch_first_True_training_True_enable_nested_tensor_True_cuda PASSED [0.0503s] [ 0%] 2024-08-06T21:15:44.6829828Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_False_use_autocast_False_d_model_12_cuda PASSED [0.0528s] [ 0%] 2024-08-06T21:15:44.6832056Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_False_use_autocast_False_d_model_256_cuda PASSED [0.0917s] [ 0%] 2024-08-06T21:15:44.6834246Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_False_use_autocast_True_d_model_12_cuda PASSED [0.1543s] [ 0%] 2024-08-06T21:15:44.6836610Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_False_use_autocast_True_d_model_256_cuda PASSED [0.2898s] [ 0%] 2024-08-06T21:15:44.6838786Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_True_use_autocast_False_d_model_12_cuda PASSED [0.0537s] [ 0%] 2024-08-06T21:15:44.6840982Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_True_use_autocast_False_d_model_256_cuda PASSED [0.0810s] [ 0%] 2024-08-06T21:15:44.6843308Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_True_use_autocast_True_d_model_12_cuda PASSED [0.1516s] [ 0%] 2024-08-06T21:15:44.6845514Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_fastpath_use_torchscript_False_enable_nested_tensor_True_use_autocast_True_d_model_256_cuda PASSED [0.2848s] [ 0%] 2024-08-06T21:15:44.6847732Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_square_input_with_no_grad_False_training_False_enable_nested_tensor_False_cuda PASSED [0.0095s] [ 0%] 2024-08-06T21:15:44.6849778Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_square_input_with_no_grad_False_training_True_enable_nested_tensor_False_cuda PASSED [0.0092s] [ 0%] 2024-08-06T21:15:44.6851831Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_square_input_with_no_grad_True_training_False_enable_nested_tensor_False_cuda PASSED [0.0084s] [ 0%] 2024-08-06T21:15:44.6853964Z test_transformers.py::TestTransformersCUDA::test_transformerencoder_square_input_with_no_grad_True_training_True_enable_nested_tensor_False_cuda PASSED [0.0090s] [ 0%] 2024-08-06T21:15:44.6855876Z test_transformers.py::TestTransformersCUDA::test_transformerencoderlayer_no_fastpath_with_hooks_nhead_3_cuda PASSED [0.0033s] [ 0%] 2024-08-06T21:15:44.6857552Z test_transformers.py::TestTransformersCUDA::test_transformerencoderlayer_no_fastpath_with_hooks_nhead_4_cuda PASSED [0.0032s] [ 0%] 2024-08-06T21:15:44.6859253Z test_transformers.py::TestTransformersCUDA::test_transformerencoderlayer_src_mask_nhead_1_cuda PASSED [0.0045s] [ 0%] 2024-08-06T21:15:44.6860911Z test_transformers.py::TestTransformersCUDA::test_transformerencoderlayer_src_mask_nhead_4_cuda PASSED [0.0042s] [ 0%] 2024-08-06T21:15:44.6862488Z test_transformers.py::TestTransformersCUDA::test_transformerencoderlayer_src_mask_nhead_8_cuda PASSED [0.0042s] [ 0%] 2024-08-06T21:15:44.6864015Z test_transformers.py::TestTransformersCUDA::test_transformerencoderlayer_subclass_cuda PASSED [0.4251s] [ 0%] 2024-08-06T21:15:44.6865694Z test_transformers.py::TestTransformersCUDA::test_transformerencoderlayer_subclass_model_cuda PASSED [1.9819s] [ 0%] 2024-08-06T21:15:44.6867263Z test_transformers.py::TestTransformersCUDA::test_with_nested_tensor_input_cuda PASSED [0.0190s] [ 0%] 2024-08-06T21:15:44.6868764Z test_transformers.py::TestSDPAFailureModesCUDA::test_dispatch_fails_no_backend_cuda PASSED [0.0021s] [ 0%] 2024-08-06T21:15:44.6870431Z test_transformers.py::TestSDPAFailureModesCUDA::test_flash_atteention_large_bf16_nan_values_cuda SKIPPED [0.0002s] (Does not support flash attention) [ 0%] 2024-08-06T21:15:44.6872417Z test_transformers.py::TestSDPAFailureModesCUDA::test_flash_attention_fail_with_non_square_causal_attention_cuda SKIPPED [0.0002s] (Does not support flash attention) [ 0%] 2024-08-06T21:15:44.6874475Z test_transformers.py::TestSDPAFailureModesCUDA::test_flash_autocast_fp32_bfloat16_cuda SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.6876379Z test_transformers.py::TestSDPAFailureModesCUDA::test_flash_autocast_fp32_float16_cuda SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.6878437Z test_transformers.py::TestSDPAFailureModesCUDA::test_flash_backward_failure_sm86plus_head_dim_193_dropout_p_0_0_cuda SKIPPED [0.0002s] (Does not support fused SDPA or not SM86+ hardware) [ 0%] 2024-08-06T21:15:44.6880785Z test_transformers.py::TestSDPAFailureModesCUDA::test_flash_backward_failure_sm86plus_head_dim_193_dropout_p_0_2_cuda SKIPPED [0.0002s] (Does not support fused SDPA or not SM86+ hardware) [ 0%] 2024-08-06T21:15:44.6883010Z test_transformers.py::TestSDPAFailureModesCUDA::test_flash_backward_failure_sm86plus_head_dim_204_dropout_p_0_0_cuda SKIPPED [0.0004s] (Does not support fused SDPA or not SM86+ hardware) [ 0%] 2024-08-06T21:15:44.6885201Z test_transformers.py::TestSDPAFailureModesCUDA::test_flash_backward_failure_sm86plus_head_dim_204_dropout_p_0_2_cuda SKIPPED [0.0002s] (Does not support fused SDPA or not SM86+ hardware) [ 0%] 2024-08-06T21:15:44.6887553Z test_transformers.py::TestSDPAFailureModesCUDA::test_flash_backward_failure_sm86plus_head_dim_256_dropout_p_0_0_cuda SKIPPED [0.0002s] (Does not support fused SDPA or not SM86+ hardware) [ 0%] 2024-08-06T21:15:44.6889705Z test_transformers.py::TestSDPAFailureModesCUDA::test_flash_backward_failure_sm86plus_head_dim_256_dropout_p_0_2_cuda SKIPPED [0.0002s] (Does not support fused SDPA or not SM86+ hardware) [ 0%] 2024-08-06T21:15:44.6892183Z test_transformers.py::TestSDPAFailureModesCUDA::test_flash_fail_fp32_cuda SKIPPED [0.0002s] (Does not support fused SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.6894063Z test_transformers.py::TestSDPAFailureModesCUDA::test_fused_kernels_nested_broadcasting_error_cases_cuda PASSED [0.0037s] [ 0%] 2024-08-06T21:15:44.6896006Z test_transformers.py::TestSDPAFailureModesCUDA::test_fused_kernels_nested_broadcasting_requires_grad_failure_cuda SKIPPED [0.0002s] (Fused SDPA was not built for this system) [ 0%] 2024-08-06T21:15:44.6897943Z test_transformers.py::TestSDPAFailureModesCUDA::test_fused_kernels_seq_len_0_inputs_fused_kernel0_cuda PASSED [0.0038s] [ 0%] 2024-08-06T21:15:44.6899743Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_fused_inputs_attn_mask_present_kernel0_cuda SKIPPED [0.0002s] (Does not support flash attention) [ 0%] 2024-08-06T21:15:44.6901550Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_fused_inputs_broadcast_kernel0_cuda PASSED [0.0019s] [ 0%] 2024-08-06T21:15:44.6903197Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_fused_inputs_dim_3_kernel0_cuda PASSED [0.0067s] [ 0%] 2024-08-06T21:15:44.6905724Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_fused_inputs_head_dim_kernel0_cuda SKIPPED [0.0002s] (Does not flash_attention fused scaled dot product attention) [ 0%] 2024-08-06T21:15:44.6907672Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_fused_inputs_invalid_dtype_kernel0_cuda PASSED [0.0016s] [ 0%] 2024-08-06T21:15:44.6909383Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_1_dimensional_inputs_kernel0_cuda PASSED [0.0014s] [ 0%] 2024-08-06T21:15:44.6911065Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_1_dimensional_inputs_kernel1_cuda PASSED [0.0014s] [ 0%] 2024-08-06T21:15:44.6912719Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_1_dimensional_inputs_kernel2_cuda PASSED [0.0014s] [ 0%] 2024-08-06T21:15:44.6914350Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_different_datatypes_kernel0_cuda PASSED [0.0014s] [ 0%] 2024-08-06T21:15:44.6916020Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_different_datatypes_kernel1_cuda PASSED [0.0015s] [ 0%] 2024-08-06T21:15:44.6917684Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_different_datatypes_kernel2_cuda PASSED [0.0014s] [ 0%] 2024-08-06T21:15:44.6919372Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_different_devices_kernel0_cuda PASSED [0.0014s] [ 0%] 2024-08-06T21:15:44.6920993Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_different_devices_kernel1_cuda PASSED [0.0015s] [ 0%] 2024-08-06T21:15:44.6922598Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_inputs_different_devices_kernel2_cuda PASSED [0.0015s] [ 0%] 2024-08-06T21:15:44.6924218Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_last_dim_stride_kernel0_cuda PASSED [0.0056s] [ 0%] 2024-08-06T21:15:44.6925783Z test_transformers.py::TestSDPAFailureModesCUDA::test_invalid_sequence_lengths_kernel0_cuda PASSED [0.0048s] [ 0%] 2024-08-06T21:15:44.6927314Z test_transformers.py::TestSDPAFailureModesCUDA::test_mem_efficient_fail_bfloat16_less_than_sm80_cuda PASSED [0.0049s] [ 0%] 2024-08-06T21:15:44.6929164Z test_transformers.py::TestSDPAFailureModesCUDA::test_nested_fails_on_padding_head_dim_cuda SKIPPED [0.0002s] (Fused SDPA was not built for this system) [ 0%] 2024-08-06T21:15:44.6931046Z test_transformers.py::TestSDPAFailureModesCUDA::test_unaligned_tensors_cuda SKIPPED [0.0002s] (Does not support fused SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.6932857Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_0_bfloat16_cuda_bfloat16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6934714Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_0_float16_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6936509Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_0_float32_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6938278Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_0_float64_cuda_float64 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6940231Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_7_bfloat16_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6941964Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_7_float16_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6943726Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_7_float32_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6945729Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_dense_dropout_0_7_float64_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6947619Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_0_bfloat16_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6949358Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_0_float16_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6951236Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_0_float32_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6953025Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_0_float64_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6954871Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_7_bfloat16_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6956708Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_7_float16_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6958452Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_7_float32_cuda_float32 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6960305Z test_transformers.py::TestSDPACUDA::test_fused_sdp_choice_cpu_type_nested_dropout_0_7_float64_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6962040Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_attention_math_with_negative_scale_kernel0_cuda PASSED [0.0023s] [ 0%] 2024-08-06T21:15:44.6964391Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6967225Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6970075Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_False_cuda_bfloat16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6972916Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6975809Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6978629Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6981519Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6984564Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6987322Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6990138Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6993097Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_False_cuda_float16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6995908Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.6998752Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7001622Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7004887Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_False_cuda_float16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7007762Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float16_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7010606Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7013408Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7016264Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_False_cuda_float32 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7019226Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7022159Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7025041Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_True_cuda_float32 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7027788Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7030586Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float32_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7033467Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7036269Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_0_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7039096Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_False_cuda_float64 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7041960Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_2_bool_mask_1_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7044695Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_False_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7047517Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_0_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7050411Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7053200Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_mask_vs_math_cpu_fused_kernel0_float64_batch_size_2_q_seq_len_267_kv_seq_len_514_n_head_3_head_dim_8_mask_dim_4_bool_mask_1_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7056070Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7058792Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7061395Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7064043Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7066664Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7069257Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7071951Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7074556Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7077247Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7079847Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7082478Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7085124Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7087759Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7090393Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7093091Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7095728Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7098446Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7101126Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7103729Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7106863Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7109582Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7112115Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7114703Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7117383Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7119965Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7122599Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7125236Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7127761Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7130378Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7133178Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7135900Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7138542Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7141112Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7143732Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7146394Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7148976Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7151561Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7154248Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7156777Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7159354Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7162073Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7164705Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7167331Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0019s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7170045Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7172669Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7175295Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7177925Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7180537Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7183213Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7185805Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7188341Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7191055Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7193681Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7196232Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7198940Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7201521Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7204442Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7207213Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7209933Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0016s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7212682Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7215292Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7217919Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7220506Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7223119Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_bfloat16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_bfloat16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7225717Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7228369Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_float16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7230978Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_float16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7233516Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7236125Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7238759Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7241305Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7243966Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7246695Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_float16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7249408Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7251960Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_float16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7254558Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7257184Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7259820Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7262416Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7265052Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_float16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7267587Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7270183Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7272813Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_float16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7275422Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7278003Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7280606Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7283265Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7285887Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7288543Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7291126Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7293782Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_float16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7296363Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7298927Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7301567Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7304635Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7307282Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7309937Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_float16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7312480Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7315067Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_float16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7317742Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7320315Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7323054Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7325890Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7328399Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_float16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7331017Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7333705Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7336310Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_float16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7338915Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7341528Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7344077Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7346818Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_float16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7349431Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7352015Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7354686Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7357318Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_float16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7359910Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7362663Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7365233Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_float16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7367770Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7370468Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7373045Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7375555Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7378252Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_float16 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7380820Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7383418Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_float16 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7386067Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7388585Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7391192Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float16_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_float16 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7393859Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7396462Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7399195Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_float32 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7401869Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_float32 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7404871Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7407578Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7410176Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7412722Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7415443Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7418053Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7420569Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_float32 SKIPPED [0.0018s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7423224Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7425822Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7428442Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_float32 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7431064Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7433631Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7436329Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7439063Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7441656Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_float32 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7444305Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7446878Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7449444Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7452040Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7454625Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7457199Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7459874Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7462457Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_float32 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7465068Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7467624Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7470275Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7472952Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_float32 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7475654Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7478247Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7480872Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7483416Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_float32 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7486012Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7488641Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7491225Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_float32 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7493859Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7496556Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7499082Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7501647Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7504750Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_float32 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7507350Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7509956Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_float32 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7512646Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7515346Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7517970Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7520579Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7523183Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7525844Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_float32 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7528358Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_float32 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7530930Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7533592Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7536158Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7538683Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7541343Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7543889Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7546428Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_float32 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7549156Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7551833Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7554454Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7557056Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7559556Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float32_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_float32 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7562187Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7564827Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7567412Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_float64 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7570078Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7572672Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7575197Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7577879Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7580491Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7583073Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7585767Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7588437Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_float64 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7591016Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7593760Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7596357Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7598970Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7601600Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7604624Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7607268Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7609940Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_float64 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7612513Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7615149Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7617750Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7620262Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7622885Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7625588Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7628250Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7630952Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_float64 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7633528Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7636037Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7638730Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7641293Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7643902Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_12_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7646518Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_False_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7649110Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_False_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7651679Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_False_cuda_float64 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7654288Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_16_causal_True_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7656890Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7659557Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_False_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7662222Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7664894Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_1_head_dim_8_causal_True_train_True_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7667536Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7670153Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_False_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7672717Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_False_cuda_float64 SKIPPED [0.0018s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7675421Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_16_causal_True_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7677996Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7680610Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_False_train_True_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7683176Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_False_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7685754Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_1030_n_head_3_head_dim_8_causal_True_train_True_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7688349Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7690977Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_False_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7693659Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_False_cuda_float64 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7696279Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_16_causal_True_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7698787Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7701443Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_False_train_True_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7704602Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7707195Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_1_head_dim_8_causal_True_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7709802Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7712475Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_False_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7714974Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_False_cuda_float64 SKIPPED [0.0015s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7717540Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_16_causal_True_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7720213Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7722801Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_False_train_True_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7725400Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_False_cuda_float64 SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7727923Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_vs_math_cpu_fused_kernel0_float64_batch_size_2_seq_len_267_n_head_3_head_dim_8_causal_True_train_True_cuda_float64 SKIPPED [0.0013s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7730004Z test_transformers.py::TestSDPACUDA::test_scaled_dot_product_fused_attention_with_inf_cuda SKIPPED [0.0012s] (Only runs on cpu) [ 0%] 2024-08-06T21:15:44.7733148Z test_transformers.py::TestSDPACUDA::test_sdp_math_gradcheck_contiguous_inputs_False_cuda SKIPPED [0.0007s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131625 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 0%] 2024-08-06T21:15:44.7737593Z test_transformers.py::TestSDPACUDA::test_sdp_math_gradcheck_contiguous_inputs_True_cuda SKIPPED [0.0010s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131255 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 0%] 2024-08-06T21:15:44.7740703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_cudnn_attention_different_dk_dv_cuda SKIPPED [0.0002s] (cuDNN Attention is not supported on this system) [ 0%] 2024-08-06T21:15:44.7742679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_cudnn_attention_fail_d128_cuda SKIPPED [0.0002s] (cuDNN Attention is not supported on this system) [ 0%] 2024-08-06T21:15:44.7745100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7747912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7750911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7753738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7756547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7759528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7762377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7765257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7768089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7770976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7773879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7776820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7779803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7782698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7785477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7788334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7791239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 0%] 2024-08-06T21:15:44.7794147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7796965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7799884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7802715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7805771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7808746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7811602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7815508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7818535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7821387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7824297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7827102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7829976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7833002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7835847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7838676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7841677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7844539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7847309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7850237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7853160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7856104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7858960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7861791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7864699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7867515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7870359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7873316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7876154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7878954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7881874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7884647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7887474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7890488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7893426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7896295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7899193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7902076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7905221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7908154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7911022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7913914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7916740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7919563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7922487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7925273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7928189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7931275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7934108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7936970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7939828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7942641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7945481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0008s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7948366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7951165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7954062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7956918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7960014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7962786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7965740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7968774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7971628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7974437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7977380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7980245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7983017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7985941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7988802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7991709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7994631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.7997467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8000369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8003256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8006504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8009407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8012248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8015075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8018053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8020908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8023680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8026621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8029445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8032319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8035204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8038106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8041115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8044024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8046881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8049817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8052713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8055524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8058493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8061287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8064084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8067028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8069861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8072759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8075634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8078536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8081489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8084356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8087216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8090122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8092988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8095851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8098784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8101532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8104654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8107606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8110439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8113317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8116359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8119314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8122116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8124987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8127816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8130719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8133540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8136343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8139255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8142065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8144886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8147828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8150642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8153603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8156534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8159299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8162166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8165054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8167898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8170785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8173714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8176541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8179413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8182265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8185080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8188009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8190939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8193976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8196848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8199639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8202534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8205749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8208587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8211513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8214376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8217169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8220075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8222936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0009s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8225743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8228760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8231784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8234663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8237468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8240303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8243198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8246074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8248908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8251800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8254650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8257447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8260328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8263192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8266069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8269082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8271936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8274777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8277576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8280404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8283276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8286119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8288943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8291872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8294781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8297546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8300454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8303385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8306592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8309518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8312361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8315248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8318062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8320904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8323815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8326694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8329546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8332452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8335222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8338035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8341036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8343934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8346710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8349839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8352711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8355457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8358361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8361218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8364067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8366914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8369797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8372686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8375453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8378357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8381337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8384225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8387044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8389988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8392861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8395631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8398576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8401392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8404554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8407477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8410288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8413168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8416086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8419055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8421963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8424773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8427615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8430561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8433344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8436134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8439036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8441882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8444790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8447671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8450492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8453397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8456345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8459154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8462083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8464935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8467748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8470698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8473460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8476308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8479217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8482022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8484905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8487751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8490600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8493583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8496450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8499294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8502185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8505428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8508271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8511190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8514033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8516822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8519764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8522602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8525500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8528494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8531386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8534318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8537185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8540007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8542893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8545731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8548539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8551379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8554226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8557048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8559966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8562803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8565768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8568660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8571490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8574362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8577239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8580067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8583000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8585840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8588609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8591527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8594407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8597247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8600178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8603140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8606452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8609254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8612087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8614972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8617803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8620634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8623518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8626386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8629167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8632074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8634906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8637706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8640760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8643702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8646447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8649397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8652239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8655116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8657983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8660875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8663786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8666562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8669401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8875770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8878522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8881552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8884403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8887100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8889809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8892510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8895239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8897948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8900663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8903339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8906394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8909123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8911789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8914460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8917263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8920090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8922766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8925437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8928139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8930819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8933503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8936178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8938856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8941554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8944251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8946932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8949599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8952361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8955111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8957762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8960436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8963121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8965792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8968459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8971135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8973844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8976531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8979205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8981884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8984588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8987359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8990146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8992898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8995588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.8998293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9001005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9003709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9006741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9009445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9012137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9014822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9017529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9020223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9023002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9025799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9028511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9031189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9033878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9036570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9039270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9041960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9044634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9047358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9050053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9052752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9055433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9058203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9060998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9063692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9066386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9069093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9071823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9074553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9077255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9079942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9082619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9085305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9087976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9090679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9093516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9096293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9098948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9101646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9104666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9107414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9110125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9112820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9115512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9118180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9120882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9123562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9126260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9129093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9131881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9134548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9137250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9139939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9142614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9145298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9148024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9150666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9153323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9156002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9158688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9161365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9164104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9166869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9169552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9172224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9174912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9177594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9180308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9183005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9185676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9188374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9191088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9193835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9196516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9199282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9202071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9205060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9207768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9210466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9213154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9215832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9218487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9221199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9223897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9226584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9229260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9231947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9234748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9237546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9240269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9242958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9245664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9248365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9251042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9253715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9256412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9259125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9261805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9264506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9267220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9269979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9272741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9275434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9278148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 1%] 2024-08-06T21:15:44.9280826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9283478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9286140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9288833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9291506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9294228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9296938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9299625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9302284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9305397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9308225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9310917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9313598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9316304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9318985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9321696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9324394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9327073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9329755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9332458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9335166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9337821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9340583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9343347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9346008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9348689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9351367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9354046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9356723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9359387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9362074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9364758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9367459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9370123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9372791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9375554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9378329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9380989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9383675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9386379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9389053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9391733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9394489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9397184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9399870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9402552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9405544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9408235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9411018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9413786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9416463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9419148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9421832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9424489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9427173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9429868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9432538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9435197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9437870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9440555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9443217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9445976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9448763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9451444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9454129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9456786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9459457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9462150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9464849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9467523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9470191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9472845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9475522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9478172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9480916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9483712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9486381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9489026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9491706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9494445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9497110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9499777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9502504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9505457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9508144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9510822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9513505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9516310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9519116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9521845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9524516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9527224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9529913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9532581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9535265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9537955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9540624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9543292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9545970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9548678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9551434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9554180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9556836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9559546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9562222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9564883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9567569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9570291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9572986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9575703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9578433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9581159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9583868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9586646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9589445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9592147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9593640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9595130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9596577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9598012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9599451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9600881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9602359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9603800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9605517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9606944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9608598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9610146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9611568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9613044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9614490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9615949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9617365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9618812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9620260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9621721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9623161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9624618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9626055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9627597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9629111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9630560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9631994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9633446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9634882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9636298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9637757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9639200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9640639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9642075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9643532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9644962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9646478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9647987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9649436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9650875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9652318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9653756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9655193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9656654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9658076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9659518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9660948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9662431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9663850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9665365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9666871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9668318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9669730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9671172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9672619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9674057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9675491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9676903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9678341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9679775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9681208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9682644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9684171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9685697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9687143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9688571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9690036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9691479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9692976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9694423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9695859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9697318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9698746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9700200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9701624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9703173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9704969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9706460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9707904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9709368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9710786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9712245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9713686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9715143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9716562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9717989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9719447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9720887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9722452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9724433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9725895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9727347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9728798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9730231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9731697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9733167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9734612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9736044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9737494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9738936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9740355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9741890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9743440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9744899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9746323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9747776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9749209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9750665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9752092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9753555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9754985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9756439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9757865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9759301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9760804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9762338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9763799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9765227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9766685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9768124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9769571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9771003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9772440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9773880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9775313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9776732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9778172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9779675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9781198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9782615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9784052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9785513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9786927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9788366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9789810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9791268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9792697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9794186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9795635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9797096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9798595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9800132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9801568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9803034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9804753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9806203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9807656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9809099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9810541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9811961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9813436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9814880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9816311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9817852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9819420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9820852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9822294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9823749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9825178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9826628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9828051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9829493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9830931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9832395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9833831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9835269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9836780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9838331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9839747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9841190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9842624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9844084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9845497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9846919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9848366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9849798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9851239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9852653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9854107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9855612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9857132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9858555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9860007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9861448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9862885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9864314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9865769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9867201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9868630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9870079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9871508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9872972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9874469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9875998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9877421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9878869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9880283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9881713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9883163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9884619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9886033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9887472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9888903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9890334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9891789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9893316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9894849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9896279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9897721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9899147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9900593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9902052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9903492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9905183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9906654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9908101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9909519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9910961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9912507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9914058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9915467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9916909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9918339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9919789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9921214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9922659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9924082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9925545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9926966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9928396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9929816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9931321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9932836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9934251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9935718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9937159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9938591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9940008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9941463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9942896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9944330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9945771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9947208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9948635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9950132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9951628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9953049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9954498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9955937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9957366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9958788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9960241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9961651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9963075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9964525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9965981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9967394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9968906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9970430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9971888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9973312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9974760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9976207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9977646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9979091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9980509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9981949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9983375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9984812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9986228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9987758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9989278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9990709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9992131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9993610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9995066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9996487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9997929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:44.9999360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0000812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0002233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0003693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0005413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0007021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0008550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0009991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0011427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0012891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0014337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0015765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0017218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0018645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0020067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0021485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0022939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0024396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0025916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0028704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0031375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0034085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0036789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0039494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0042213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0044922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0047587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0050265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0052962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0055744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0058521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0061319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0064126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0066826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0069533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0072237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0074927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0077622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0080319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0082970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0085671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0088375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0091069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0093782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0096547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0099328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0102016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0104977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0107685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0110378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0113083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0115779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0118451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0121151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0123840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0126521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0129208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0132031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0134818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0137492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 2%] 2024-08-06T21:15:45.0140190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0142875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0145540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0148201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0150856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0153563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0156230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0158922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0161600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0164279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0167020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0169759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0172460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0175171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0177873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0180557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0183249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0185970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0188663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0191370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0194114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0196836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0199525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0202270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0205371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0208071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0210775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0213476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0216153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0218836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0221532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0224202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0226894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0229624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0232338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0235015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0237797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0240630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0243323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0246012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0248734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0251439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0254143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0256823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0259538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0262247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0264945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0267652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0270339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0273092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0275983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0278796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0281487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0284199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0286943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0289721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0292464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0295201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0297882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0300566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0303270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0306324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0309141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0311919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0314607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0317326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0320024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0322883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0325595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0328299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0330982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0333680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0336359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0339128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0341907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0344647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0347404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0350088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0352763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0355434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0358082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0360825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0363510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0366166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0368850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0371562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0374244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0376928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0379757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0382558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0385255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0387971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0390672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0393402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0396105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0398823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0401513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0404494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0407234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0409895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0412569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0415391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0418202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0420880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0423575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0426264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0428952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0431638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0434353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0437057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0439742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0442408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0445083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0447780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0450543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0453338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0456029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0458729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0461431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0464107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0466788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0469480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0472155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0474819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0477482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0480201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0482872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0485618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0488409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0491105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0493808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0496490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0499165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0501862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0504810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0507518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0510225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0512925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0515623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0518300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0521100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0523916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0526605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0529296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0531983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0534693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0537365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0540029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0542702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0545370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0548068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0550741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0553410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0556175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0558940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0561607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0564288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0567014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0569694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0572358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0575048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0577752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0580427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0583107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0585816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0588533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0591301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0594104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0596765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0599449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0602119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0605070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0607820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0610509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0613178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0615857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0618535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0621229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0623914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0626697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0629486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0632171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0634877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0637553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0640234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0642941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0645629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0648288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0650965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0653667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0656351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0659026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0661806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0664546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0667209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0669867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0672542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0675225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0677905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0680552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0683228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0685910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0688607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0691276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0693987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0696757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0699521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0702184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0705155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0707885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0710582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0713236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0715968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0718678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0721363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0724043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0726727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0729406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0732160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0734960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0737637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0740327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0743026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0745697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0748350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0751045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0753756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0756419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0759108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0761828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0764484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0767241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0770028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0772749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0775435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0778116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0780792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0783493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0786183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0788883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0791574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0794293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0796970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0799609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0802356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0805407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0808091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0810771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0813449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0816155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0818825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0821491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0824176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0826863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0829561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0832216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0834898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0837700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0840496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0843196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0845877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0848575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0851269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0853934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0856618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0859307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0861966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0864628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0867307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0870019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0872695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0875434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0878202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0880887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0883546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0886220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0888890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0891574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0894287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0896969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0899632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0902326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0905309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0908031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0910832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0913633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0916288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0918962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0921645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0924340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0927001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0929652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0932294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0934963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0937621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0940284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0942946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0945706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0948445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0951103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0953782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0956492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0959185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0961860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0964547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0967226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0969910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0972585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0975280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0978019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0980781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0983530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0986210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0988902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0991575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0994278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0996980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.0999667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1002337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1005310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1008008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1010705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1013380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1016170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1018974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1021666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1024342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1027033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1029717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1032442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1035134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1037804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1040488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1043187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1045872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1048542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1051295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1054086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1056750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1059438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1062115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1064803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1067454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1070117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1072820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1075501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1078195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1080868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1083518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1086293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1089044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1091727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1094455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1097147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1099805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1102463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1105427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1108133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1110820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1113512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1116187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1118838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1121606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1124362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1127024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1129720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1132383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1135023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1137688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1140390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1143060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1145708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1148404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1151087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1153765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1156511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1159295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1161990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1164672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1167367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1170054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1172744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1175433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1178112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1180784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1183482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1186157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1188814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1191595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1194453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1197128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1199792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1202464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1205445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1208164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1210835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1213532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1216211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1218866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1221560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1224233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1226925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1229722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1232497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1235150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1237841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1240557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1243220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1245898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1248587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1251226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1253873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1256540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1259237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1261894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1265333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1268126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1270791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1273456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1276134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1278822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1281502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1284177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1286814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1289487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1292178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1294903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1297587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1300358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1303138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1306088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1308944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1311640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1314304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1316985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1319622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1322290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1324963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1327617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1330297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1332959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1335774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1338551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1341200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1343863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1346558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1349219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1351882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1354572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1357279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1359945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1362616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1365301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 3%] 2024-08-06T21:15:45.1367983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1370717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1373487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1376150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1378827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1381488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1384156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1385568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1386984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1388435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1389840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1391271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1392692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1394146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1395628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1397141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1398578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1399997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1401405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1402835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1404543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1405983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1407400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1408832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1410270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1411681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1413104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1414629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1416173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1417564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1418993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1420402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1421839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1423234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1424653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1426067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1427478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1428912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1430316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1431750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1433243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1434757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1436159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1437593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1439036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1440468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1441878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1443319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1444749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1446171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1447588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1449025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1450452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1451842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1453330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1454826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1456257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1457656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1459096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1460506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1461940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1463345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1464763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1466195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1467642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1469090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1470512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1472038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1473583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1475025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1476455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1477920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1479362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1480802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1482233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1483674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1485102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1486538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1487968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1489395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1490920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1492423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1493899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1495335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1496795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1498230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1499668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1501104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1502554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1504260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1505774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1507213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1508686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1510224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1511764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1513221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1514665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1516113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1517539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1519013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1520452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1521879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1523293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1524745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1526178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1527610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1529126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1530653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1532093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1533506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1534946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1547103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1548726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1550190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1551632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1553064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1554520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1555933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1557369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1558943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1560529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1561944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1563383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1564807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1566247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1567654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1569091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1570525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1571951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1573384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1574798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1576234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1577738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1579273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1580684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1582139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1583579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1585017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1586441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1587897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1589347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1590773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1592221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1593704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1595164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1596660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1598181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1599624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1601072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1602486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1604237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1605703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1607165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1608579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1610043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1611474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1612903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1614333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1615897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1617459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1618898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1620363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1621790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1623238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1624678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1626121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1627544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1628993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1630457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1631890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1633310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1634822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1636360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1637766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1639200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1640649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1642099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1643516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1644954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1646376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1647827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1649243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1650693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1652111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1653627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1655129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1656548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1658000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1659442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1660874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1662295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1663746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1665177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1666610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1668037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1669476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1670896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1672393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1673890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1675308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1676752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1678171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1679614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1681040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1682488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1683894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1685319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1686750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1688207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1689646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1691160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1692681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1694183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1695616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1697047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1698500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1699969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1701420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1702843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1704564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1706019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1707451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1708860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1710449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1711989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1713424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1714844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1716297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1717732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1719151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1720611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1722032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1723478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1724890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1726330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1727754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1729276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1730795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1732230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1733660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1735114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1736523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1737963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1739386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1740821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1742240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1743653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1745104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1746525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1748027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1749530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1750979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1752403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1753838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1755248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1756692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1758124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1759550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1760983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1762410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1763871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1765286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1766792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1768315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1769774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1771194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1772641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1774057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1775498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1776908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1778334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1779772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1781199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1782633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1784041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1785558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1787073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1788495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1789930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1791378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1792803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1794276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1795696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1797138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1798565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1800023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1801445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1802871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1804931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1806787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1808231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1809655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1811128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1812536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1813972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1815400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1816840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1818244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1819675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1821124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1822550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1824061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1825559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1826997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1828427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1829862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1831298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1832737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1834166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1835601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1837016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1838457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1839885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1841332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1842816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1844327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1845765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1847171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1848597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1850011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1851467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1852876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1854309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1855723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1857161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0007s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1858569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1859993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1861501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1863030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1864437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1865854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1867308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1868737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0005s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1870188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1871614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1873060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1874493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1875924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1877344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1878782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1880527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1882064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1883473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1884896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1886345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1887749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1889174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1890614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1892064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1893500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1894939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1896371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1897817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1899305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1900851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1902279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1903738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1905540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1906969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1908410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1909849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1911314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1912733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1914175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1915604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1917029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1918569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1920127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1921577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1923010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1924432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1925869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1927292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1928712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1930154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1931601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1933048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1934464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1935895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1937397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1938936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1940353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1941812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1943246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1944705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1946127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1947553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1948999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1950427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1951870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1953290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1954729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1956230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1957745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1959158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1960599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1962047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1963483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1964898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1966344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1967781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1969188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1970617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1972059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1973504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1974990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1976506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1977930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1979369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1980789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1982225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1983637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1985074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1986479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1987898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1989316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1990750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1992173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1993707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1995240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1996659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1998075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.1999483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2000950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2002384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2003802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2005647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2007097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2008538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2009961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2011415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2012954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2014513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2015933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2017369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2018794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2020243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2021672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2023104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2024529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2025973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2027393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2028832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2030253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2031701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2033204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2034702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2036143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2037581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2039018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2040430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2041902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2043367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2044784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 4%] 2024-08-06T21:15:45.2046223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2047659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2049094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2050531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2052043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2053580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2055004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2056428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2057843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2059276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2060704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2062154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2063564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2065003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2066435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2067859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2069269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2070761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2072299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2073705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2075132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2076556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2077993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2079402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2080835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2082269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2083710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2085121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2086538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2087946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2089433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2090942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2092347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2093814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2095249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2096670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2098078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2099509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2100928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2102337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2103748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2105516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2106994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2108535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2110059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2111499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2112953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2114376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2115799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2117226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2118671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2120085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2121539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2122965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2124409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2125813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2127311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2128834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2130255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2131701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2133126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2134560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2135983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2137410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2138821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2140253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2141703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2143130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2144533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2146043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2147546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2148974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2150392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2151842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2153276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2154684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2156112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2157528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2158953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2160351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2161795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2163207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2164718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2166196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2167617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2169036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2170477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2171901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2173318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2174759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2176183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2177601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2179015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2180774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2182244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2183761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2185267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2186701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2188120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2189548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2190959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2192378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2193849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2195257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2196672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2198084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2199520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2200915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2202427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2204162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2205626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2207030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2208459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2209872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2211300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2212719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2214123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2215552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2216980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2218407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2219814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2221397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2222929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2224353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2225764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2227196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2228604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2230011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2231423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2232845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2234270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2235677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2237107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2238515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2240019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2241515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2242937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2244348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2245782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2247190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2248607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2250031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2251494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2252887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2254297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2255736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2257154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2258646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2260136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2261577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2262983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2264392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2265797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2267221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2268634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2270060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2271481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2272897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2274332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2275724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2277214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2279360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2280796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2282223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2283649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2285076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2286511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2287918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2289347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2290765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2292231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2293688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2295108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2296530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2298016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2299502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2300900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2302352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2303780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2305441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2306847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2308281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2309699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2311118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2312553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2313990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2315416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2316933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2318474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2319895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2321333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2322766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2324188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2325616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2327056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2328461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2329895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2331316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2332752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2334145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2335624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2337156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2338577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2339998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2341413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2342860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2344281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2345699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2347098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2348529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2349955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2351385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2352804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2354310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2355826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2357237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2358670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2360100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2361543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2362963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2364401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2365814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2367249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2368656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2370084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2371503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2373022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2374517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2375925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2377361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2378786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2380209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2381627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2383066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2384482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2385898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2387307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2388750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2390175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2391681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2393183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2394660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2396094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2397508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2398935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2400339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2401789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2403189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2404846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2406321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2407769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2409164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2410689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2412248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2413686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2415089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2416502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2417941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2419363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2420799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2422241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2423680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2425113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2426549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2427962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2429476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2430988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2432444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2433868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2435314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2436734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2438138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2439570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2441000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2442466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2443880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2445321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2446739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2448244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2449734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2451162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2452603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2454049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2455468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2456897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2458329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2459768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2461197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2462638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2464088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2465515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2467015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2468521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2469956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2471371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2472808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2474219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2475656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2477082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2478489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2479915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2481337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2482799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2484203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2485718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2487220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2488660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2490060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2491486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2492926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2494404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2495810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2497243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2498657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2500079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2501505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2502925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2504609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2506168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2507708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2509111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2510546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2511968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2513394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2514803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2516239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2517650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2519054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2520485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2521907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2523349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2524836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2526356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2527782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2529227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2530652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2532090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2533517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2534971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2536382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2537821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2539243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2540661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2542091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2543574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2545098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2546521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2547961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2549376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2550809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2552255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2553687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2555097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2556526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2557950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2559362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2560785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2562314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2563850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2565262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2566696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2568119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2569559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2570965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2572415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2573820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2575248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2576646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2578070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2579480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2580966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2582491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2583897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2585327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2586748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2588166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2589571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2591007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2592459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2593919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2595330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2596771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2598188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2599664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2601173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2602613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2604301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2605737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2607164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2608572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2610008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2611411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2612854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2614266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2615701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2617101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2618629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2620160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2621581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2623022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2624432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2625861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2627279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2628699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2630110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2631543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2632987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2634412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2635823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2637337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2638848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2640255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2641680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2643115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2644556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2645951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2647374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2648792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2650220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2651623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2653067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2654477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2655981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2657469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2658895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2660308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2661728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2663167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2664566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2665997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2667420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2668839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2670250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2671692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2673128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2674543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2676029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2677535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2678948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2680345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2681764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2683185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2684617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2686021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2687437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2688848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2690283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2691679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2693110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2694634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2696177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2697577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2699005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 5%] 2024-08-06T21:15:45.2700430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2701852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2703289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2704968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2706458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2707887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2709320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2710737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2712160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2713697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2715221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2716620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2718050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2719472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2720868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2722291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2723709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2725143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2726538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2727959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2729395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2730848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2732346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2733872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2735301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2736765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2738194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2739640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2741081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2742532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2743972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2745392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2746836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2748270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2749692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2751180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2752735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2754166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2755605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2757031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2758478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2759913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2761356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2762796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2764229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2765684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2767109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2768545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2770068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2771604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2773046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2774487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2775928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2777383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2778804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2780250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2781668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2783119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2784546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2785967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2787412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2788918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2790436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2791854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2793316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2794797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2796231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2797651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2799102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2800529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2801955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2803386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2805053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2806515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2808043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2809590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2811022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2812484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2813911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2815343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2816764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2818208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2819608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2821035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2822464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2823914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2825321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2826808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2828329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2829755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2831185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2832622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2834072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2835503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2836946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2838371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2839821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2841262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2842727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2844151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2845692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2847270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2848694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2850134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2851566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2853041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2854448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2855887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2857319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2858768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2860186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2861641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2863060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2864581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2866082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2867498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2868946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2870389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2871832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2873248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2874700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2876143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2877582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2879005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2880470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2881906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2883411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2884917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2886355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2887787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2889201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2890639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2892061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2893515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2894975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2896412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2897840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2899291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2900721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2902233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2903752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2905454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2906874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2908316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2909740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2911191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2912627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2914049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2915493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2916928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2918368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2919783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2921348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2922880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2924299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2925710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2927156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2928579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2930001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2931434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2932860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2934305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2935717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2937153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2938581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2940107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2941632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2943072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2944511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2945969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2947392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2948832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2950264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2951729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2953162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2954590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2956040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2957464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2958969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2960475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2961936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2963371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2964811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2966227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2967666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2969099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2970541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2971963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2973389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2974839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2976247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2977752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2979279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2980741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2982166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2983611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2985038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2986484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2987899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2989338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2990773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2992217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2993627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2995070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2996584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2998092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.2999512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3000949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3002398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3003821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3005495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3006950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3008402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3009827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3011279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3012704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3014125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3015680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3017205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3018644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3020077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3021555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3022968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3024399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3025814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3027261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3028664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3030096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3031543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3032986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3034469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3035968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3037404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3038834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3040269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3041701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3043136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3044566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3045998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3047414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3048859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3050299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3051760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3053264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3054795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3056230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3057645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3059083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3060498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3061952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3063356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3064791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3066214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3067655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3069067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3070496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3071992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3073514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3074916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3076341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3077767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3079199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3080645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3082057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3083503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3084932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3086363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3087781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3089222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3090726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3092238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3093649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3095115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3096538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3097945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3099364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3100803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3102253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3103658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3105344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3106775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3108214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3109723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3111282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3112707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3114154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3115569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3117008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3118429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3119865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3121326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3122745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3124191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3125628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3127061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3128565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3130084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3131526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3132947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3134359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3135802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3137224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3138635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3140068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3141510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3142955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3144363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3145790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3147214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3148777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3150270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3151719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3153143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3154596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3156006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3157444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3158878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3160308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3161756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3163178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3164608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3166022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3167517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3169007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3170441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3171879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3173312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3174720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3176157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3177582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3179004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3180422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3181862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3183305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3184716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3186214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3187732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3189183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3190603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3192044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3193465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3194953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3196376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3197808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3199227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3200681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3202086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3203491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3205268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3206820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3208247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3209662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3211121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3212544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3213976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3215388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3216822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3218242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3219666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3221090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3222508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3224025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3225511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3226939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3228359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3229804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3231233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3232666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3234082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3235509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3236902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3238319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3239731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3241184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3243232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3244751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3246179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3247598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3249019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3250422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3251888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3253319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3254741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3256148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3257591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3259026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3260453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3261966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3263488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3264936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3266354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3267793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3269208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3270647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3272077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3273499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3274913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3276352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3277763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3279191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3280680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3282218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3283627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3285043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3286489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3287914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3289345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3290762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3292212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3293640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3295116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3296537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3297977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3299480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3301018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3302434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3303871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3305578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3307018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3308447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3309874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3311325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3312733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3314167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3315589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3317022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3318539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3320077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3321517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3322959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3324369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3325774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3327219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3328646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3330072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3331501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3332941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3334366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3335792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3337286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3338802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3340210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3341654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3343056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3344484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3345896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3347299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3348725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3350135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3351596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3353004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3354427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3355923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3357460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 6%] 2024-08-06T21:15:45.3358866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3360297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3361748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3363192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3364600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3366035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3367465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3368895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3370328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3371766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3373201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3374698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3376199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3377605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3379042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3380466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3381909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3383313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3384756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3386187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3387588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3389013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3390434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3391887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3393363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3394909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3396329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3397771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3399182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3400611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3402039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3403481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3405123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3406570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3407980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3409400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3410809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3412337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3413877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3415291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3416705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3418114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3419542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3420981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3422406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3423817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3425247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3426671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3428105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3429509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3431015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3432539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3433948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3435374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3436792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3438241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3439643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3441089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3442500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3443930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3445328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3446750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3448161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3449693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3451214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3452620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3454049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3455465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3456890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3458293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3459728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3461172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3462591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3463998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3465436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3466857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3468355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3469865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3471308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3472748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3474182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3475622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3477030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3478466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3479868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3481307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3482722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3484156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3485551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3486967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3488449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3489944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3491378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3492784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3494259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3495677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3497099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3498508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3499939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3501381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3502804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3504458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3505909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3507436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3508964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3510369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3511797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3513232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3514623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3516040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3517450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3518882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3520273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3521711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3523127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3524557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3526028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3527534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3528946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3530371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3531818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3533221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3534660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3536091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3537513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3538921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3540361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3541805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3543226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3544706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3546245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3547655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3549070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3550475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3551902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3553341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3554763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3556185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3557592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3559032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3560436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3561856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3563358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3564906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3566325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3567767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3569210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3570673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3572100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3573530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3574973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3576408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3577851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3579280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3580740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3582246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3583756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3585167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3586615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3588051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3589482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3590922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3592359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3593804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3595258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3596698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3598132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3599585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3601101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3602626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3604303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3605816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3607256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3608693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3610128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3611610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3613030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3614449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3615890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3617324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3618751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3620282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3621853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3623286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3624723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3626147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3627588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3629021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3630457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3631888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3633325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3634752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3636167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3637600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3639102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3640631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3642053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3643492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3644927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3646371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3647786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3649226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3650654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3652095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3653500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3654932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3656352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3657851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3659363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3660784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3662225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3663656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3665081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3666490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3667939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3669376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3670822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3672247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3673703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3675137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3676631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3678161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3679595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3681084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3682520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3683961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3685387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3686838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3688248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3689682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3691135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3692592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3694042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3695569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3697078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3698510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3699949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3701396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3702848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3704545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3706007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3707427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3708876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3710317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3711780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3713204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3714776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3716337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3717775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3719201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3720630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3722092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3723504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3724947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3726372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3727814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3729232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3730677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3732111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3733633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3735173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3736605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3738029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3739481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3740912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3742324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3743768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3745207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3746633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3748048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3749499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3750941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3752450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3753967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3755400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3756821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3758249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3759659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3761099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3762546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3763961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3765387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3766807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3768258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3769664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3771190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3772712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3774165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3775589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3777036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3778469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3779922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3781373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3782806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3784252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3785691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3787131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3788553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3790062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3791594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3793018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3794473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3795929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3797359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3798785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3800207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3801679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3803110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3804772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3806236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3807664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3809216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3810737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3812186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3813620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3815075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3816495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3817926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3819351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3820804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3822231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3823668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3825091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3826512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3828010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3829520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3830965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3832393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3833826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3835235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3836671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3838094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3839523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3840941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3842385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3843815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3845224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3846728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3848235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3849683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3851110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3852558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3853984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3855433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3856853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3858295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3859706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3861164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3862574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3863997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3865494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3867008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3868432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3869847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3871311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3872738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3874162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3875576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3877018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3878443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3879875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3881320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3882759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3884264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3885783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3887199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3888626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3890084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3891519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3892947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3894404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3895855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3897257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3898683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3900106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3901568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3903060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3904825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3906296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3907721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3909150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3910570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3912023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3913450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3914882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3916289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3917726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3919160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3920584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3922126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3923673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3925097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3926526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3927940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3929358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3930798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3932221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3933653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3935069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3936506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3937916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3939339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3940833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3942368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3943776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3945201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3946629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3948062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3949496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3950923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3952373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3953810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3955242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3956665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3958106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3959614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3961138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3962553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3963985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3965411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3966841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3968250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3969675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3971144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3972552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3973986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3975408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3976843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3978335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3979853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3981294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3982734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3984142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3985582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3987007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3988460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3989878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3991313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3992761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3994229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3995667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3997162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.3998682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.4000097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.4001532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.4002941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.4004632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.4006068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.4007494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.4008910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.4010343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.4011788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.4013200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 7%] 2024-08-06T21:15:45.4014626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4016046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4017615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4019131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4020558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4022005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4023457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4024869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4026307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4027739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4029191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4030607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4032041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4033468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4034882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4036370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4037926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4039362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4040785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4042223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4043632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4045064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4046493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4047912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4049322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4050780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4052200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4053597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4055090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4056592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4058032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4059456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4060908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4062329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4063773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4065183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4066607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4068014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4069453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4070864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4072280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4073768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4075266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4076686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4078095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4079532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4080967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4082391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4083802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4085241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4086661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4088093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4089510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4090972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4092476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4094005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4095452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4096879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4098331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4099744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4101203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4102629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4104293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4105711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4107143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4108565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4110003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4111546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4113103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4114524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4115951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4117378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4118786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4120226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4121680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4123109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4124525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4125968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4127405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4128838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4130326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4131873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4133300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4134731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4136152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4137566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4138998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4140406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4141844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4143261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4144705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4146117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4147539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4149026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4150547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4151965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4153405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4154830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4156246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4157668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4159080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4160509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4161931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4163357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4164773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4166205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4167695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4169212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4170629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4172058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4173478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4174891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4176291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4177708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4179143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4180553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4181981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4183399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4184829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4186316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4188492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4189915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4191382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4192800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4194276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4195701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4197134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4198575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4199993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4201459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4202896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4204593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4206198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4207756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4209176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4210596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4212030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4213470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4214888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4216317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4217733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4219148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4220590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4222024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4223456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4224945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4226464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4227865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4229298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4230719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4232181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4233583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4235018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4236442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4237876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4239286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4240700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4242145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4243628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4245128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4246523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4247956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4249379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4250797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4252221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4253651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4255072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4256483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4257891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4259313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4260750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4262257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4263766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4265181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4266620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4268027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4269455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4270866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4272325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4273734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4275160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4276567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4278002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4279397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4280872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4282388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4283805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4285220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4286624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4288059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4289473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4290910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4292318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4293748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4295217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4296642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4298045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4299540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4301069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4302475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4303893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4305567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4307020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4308424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4309855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4311285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4312713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4314111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4315535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4316946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4318379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4319915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4321467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4322906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4324330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4325757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4327166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4328601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4330023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4331461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4332871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4334310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4335727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4337145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4338677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4340203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4341647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4343055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4344490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4345897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4347330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4348726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4350138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4351572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4353004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4354393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4355806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4357286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4358801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4360193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4361617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4363060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4364476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4365898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4367311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4368750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4370172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4371618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4373035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4374463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4375955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4377462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4378868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4380298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4395965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4397740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4399196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4400639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4402097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4403552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4405350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4406913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4408336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4410000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4411545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4412989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4414414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4415851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4417283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4418733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4420171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4421602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4423067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4424892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4426453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4428196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4429933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4431495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4433051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4434581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4436212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4437674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4439207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4445234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4446788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4448203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4449625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4451026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4452435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4453957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4455471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4456865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4458273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4459703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4461117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4462576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4464000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4465424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4466837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4468240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4469644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4471047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4472549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4474029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4475417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4476822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4478245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4479640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4481041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4482472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4483899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4485285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4486689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4488110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4489516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4490999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4492531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4493962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4495368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4496784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4498188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4499613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4501029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4502436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4503852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4506708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4508152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4509538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4510945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4512503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4514076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4515475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4516894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4518306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4519736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4521134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4522535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4523976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4525395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4526820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4528226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4529659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4531157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4532668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4534098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4535540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4536973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4538393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4539803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4541232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4542641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4544055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4545474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4546890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4548326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4549815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4551326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4552734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4554198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4555603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4557021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4558434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4559873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4561272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4562672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4564130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4565556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4566975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4568460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4569984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4571404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4572834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4574272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4575697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4577108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4578524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4579920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4581346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4582766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4584197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4585626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4587112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4588638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4590039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4591467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4592924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4594381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4595776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4597196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4598613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4600049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4601450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4602862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4604563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4606133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4607676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4609079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4610496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4611904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4613316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4614733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4616162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4617576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4618982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4620381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4621812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4623222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4624713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4626219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4627633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4629071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4630475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4631912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4633323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4634787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4636198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4637622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4639047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4640488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4641890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4643377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4644962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4646380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4647790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4649192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4650634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4652045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4653466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4654890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4656323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4657740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4659163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4660562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4661984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4663471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4664967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4666386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4667801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4669244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4670642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4672065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4673479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4674928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4676330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4677759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4679158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4680579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4682044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4683534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4684973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4686388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4687808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4689203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 8%] 2024-08-06T21:15:45.4690625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4692039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4693500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4694913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4696346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4697766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4699179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4700654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4702149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4703581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4705289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4706744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4708160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4709591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4710992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4712415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4713816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4715260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4716652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4718055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4719579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4721116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4722506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4723907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4725348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4726756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4728164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4729563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4730996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4732405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4733816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4735235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4736661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4738147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4739654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4741060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4742468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4743914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4745375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4746807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4748216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4749645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4751033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4752446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4753867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4755294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4756775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4758271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4759671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4761095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4762494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4763914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4765338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4766749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4768168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4769556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4770983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4772398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4773838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4775237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4776741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4778239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4779654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4781056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4782458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4783901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4785324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4786737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4788133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4789550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4790935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4792385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4793812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4795315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4796784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4798192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4799600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4801016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4802426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4803849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4805558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4806974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4808384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4809793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4811229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4812643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4814180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4815691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4817106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4818514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4819925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4821313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4822712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4824154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4825553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4826961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4828372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4829805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4831195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4832697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4834217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4835669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4837086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4838521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4839944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4841374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4842812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4844253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4845698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4847132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4848571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4849989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4851500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4853005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4854445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4855860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4857302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4858721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4860140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4861559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4862983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4864448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4865862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4867298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4868722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4870234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4871728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4873158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4874605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4876055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4877471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4878906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4880336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4881770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4883198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4884636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4886076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4887496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4888992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4890495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4891933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4893389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4894840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4896259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4897692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4899115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4900544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4901957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4903373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4905080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4906500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4908040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4909573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4911012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4912417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4913854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4915291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4916726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4918137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4919575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4920981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4922415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4923822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4925223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4926733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4928240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4929656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4931058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4932490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4933941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4935356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4936802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4938233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4939677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4941092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4942509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4943968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4945561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4947103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4948529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4949970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4951409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4952850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4954288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4955720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4957144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4958565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4959975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4961398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4962834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4964259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4965770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4967288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4968730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4970141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4971577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4972997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4974465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4975879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4977307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4978737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4980196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4981606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4983022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4984564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4986074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4987503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4988926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4990366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4991784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4993251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4994696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4996138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4997566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.4998995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5000397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5001829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5003330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5005121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5006570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5007991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5009436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5010833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5012260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5013683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5015137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5016541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5017974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5019386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5020819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5022352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5025204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5027909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5030579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5033227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5035892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5038593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5041280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5043978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5046707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5049357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5052029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5054681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5057420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5060207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5062901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5065563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5068243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5070926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5073631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5076322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5078995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5081692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5084377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5087086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5089762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5092551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5095327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5097987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5100638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5103318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5106309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5108995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5111663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5114358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5117024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5119718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5122390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5125059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5127850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5130622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5133276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5135951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5138656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5141339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5144006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5146686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5149364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5152009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5154674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5157358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5160046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5162770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5165497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5168156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5170834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5173501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5176178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5178843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5181525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5184163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5186830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5189495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5192179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5194922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5197669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5200407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5203092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5206065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5208746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5211416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5214145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5216825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5219477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5222146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5224814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5227471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5230114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5232920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5235686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5238350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5241029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5243702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5246376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5249031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5251683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5254357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5257025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5259698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5262354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5265022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5267790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5270550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5273204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5275879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5278566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5281236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5283878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5286538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5289206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5291857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5294568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5297233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5299896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5303296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5306359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5309024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5311703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5314369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5317031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5319674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5322353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5325030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5327686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5330360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5333068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5335719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5338490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5341295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5343986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5346707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5349375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5352019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5354677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5357331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5359979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5362646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5365319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5367995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5370617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5373348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5376101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5378746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5381404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5384067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5386732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5389386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5392038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5394776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5397457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5400151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5402800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5405724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5408408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5411190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5413977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5416639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5419310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5421986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5424621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5427281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5429964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5432618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5435252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5437947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5440634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5443290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5446064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5448848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5451529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5454181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5456879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5459555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5462248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5464928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5467595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5470259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5472945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5475658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5478315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5481054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5483805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5486433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5489087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5491758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5494500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5497167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5499825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5502488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5505400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5508072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5510733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5513419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5516209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5518982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5521631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5524305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5527001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5529684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5532366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5535040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5537698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5540366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5543030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5545701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5548386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5551121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5553823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5556504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5559184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5561845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5564497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5567167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5569835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5572494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5575164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5577835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5580510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5583172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5585906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5588648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5591321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5594032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5596692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5599356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5602054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5604951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5607624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5610290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5612954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5615601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5618233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5620967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5623748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 9%] 2024-08-06T21:15:45.5626395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5629052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5631708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5634368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5637012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5639649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5642323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5645002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5647724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5650388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5653043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5655821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5658569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5661246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5663942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5666630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5669284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5671944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5674613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5677277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5679949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5682626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5685297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5687959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5690708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5693476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5696149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5698851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5701518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5704413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5707111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5709820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5712486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5715143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5717858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5720526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5723198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5725981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5728764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5731447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5734113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5736808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5739475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5742142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5744796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5747447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5750109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5752784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5755453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5758105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5760841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5763623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5766289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5768942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5771592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5774291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5776940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5779591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5782258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5784936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5787585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5790269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5792982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5795669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5798405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5801158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5803791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5806690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5809350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5812006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5814666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5817362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5819985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5822628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5825282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5827943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5830599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5833356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5836141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5838803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5841472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5844154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5846830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5849536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5852210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5854857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5857607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5860300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5862965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5865647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5868386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5871140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5873785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5876428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5879101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5881774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5884442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5887086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5889745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5892421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5895133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5897816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5900474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5903213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5906189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5908838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5911504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5914188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5916842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5919503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5922173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5924880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5927540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5930204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5932877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5935537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5938270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5941025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5943681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5946392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5949057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5951726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5954368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5957035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5959701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5962340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5964992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5967664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5970298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5973021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5975828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5978546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5981214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5983873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5986522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5989192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5991854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5994546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5997201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.5999866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6002515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6005394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6008161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6010938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6013600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6016241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6018895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6021535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6024179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6026820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6029488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6032177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6034837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6037474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6040154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6042904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6045652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6048322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6050988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6053650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6056313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6058992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6061658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6064325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6066997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6069616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6072270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6074932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6077658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6080374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6083020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6085716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6088354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6091001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6093709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6096380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6099033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6101672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6104565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6107255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6109909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6112549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6115319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6118088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6120737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6123403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6126053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6128702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6131337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6133959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6136582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6139236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6141905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6144535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6147206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6149938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6152647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6155280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6157935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6160634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6163286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6165933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6168582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6171257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6173937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6176589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6179280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6181962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6184690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6187407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6190056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6192757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6195419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6198074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6200755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6203409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6206308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6208954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6211613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6214280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6216947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6219674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6222435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6225104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6227792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6230454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6233114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6235805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6238466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6241127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6243790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6246528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6249208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6251846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6254636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6257383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6260032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6262755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6265415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6268082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6270714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6273381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6276043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6278701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6281381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6284048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6286687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6289439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6292198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6294888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6297557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6300252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6302906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6305821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6308529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6311242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6313913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6316577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6319227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6321891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6324645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6327390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6330058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6332722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6335390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6338046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6340705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6343390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6346054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6348710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6351370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6354021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6356668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6359379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6362148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6364815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6367482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6370132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6371570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6372988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6374408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6375823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6377239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6378647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6380058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6381449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6382871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6384358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6385835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6387247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6388657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6390092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6391480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6392932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6394365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6395794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6397200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6398629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6400066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6401500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6402984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6404730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6406175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6407613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6409053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6410460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6411889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6413307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6414715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6416114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6417556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6418978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6420393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6421906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6423444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6424858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6426260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6427702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6429124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6430559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6431967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6433391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6434807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6436246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6437667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6439097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6440586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6442117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6443524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6444955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6446372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6447803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6449212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6450616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6452056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6453474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6454898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6456307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6457748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6459236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6460771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6462170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6463601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6465018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6466414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6467851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6469265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6470703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6472098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6473521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6474936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6476367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6477859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6479368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6480766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6482183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6483572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6484972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6486389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6487824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6489240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6490634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6492057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6493510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6494926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6496397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6497937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6499357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6500775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6502186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6503633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6505296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6506723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6508183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6509605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6511046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6512459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6513883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6515289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6516834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6518361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6519776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6521191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6522635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6524033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6525435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6526864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6528305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6529729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6531136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6532556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6533961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6535444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6536921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6538366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6539780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6541201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6542602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6544033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6545502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6546907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6548346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6549754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6551181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6552572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6554057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6555541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6556962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6558379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6559800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6561205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6562632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6564030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6565424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6566847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6568275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6569690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6571087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6572589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6574094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6575509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6576914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6578362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6579773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6581185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6582589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6584008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6585411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6586801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6588234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6589637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6591150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6592628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6594084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6595487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6596915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6598332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6599746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6601154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6602589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6604239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6605676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6607114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6608543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6610060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6611564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6612991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6614402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6615817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6617222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6618651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6620055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6621467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 10%] 2024-08-06T21:15:45.6622856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6624264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6625692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6627080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6628505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6629989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6631502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6632897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6634310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6635720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6637142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6638550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6639965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6641368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6642798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6644196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6645593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6647018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6648515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6650021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6651417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6652832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6654244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6655643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6657037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6658473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6659874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6661279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6662678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6664082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6665506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6666967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6668473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6669879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6671311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6672714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6674124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6675530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6676965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6678381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6679807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6681220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6682641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6684054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6685524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6687024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6688442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6689848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6691242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6692664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6694107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6695520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6696919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6698349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6699762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6701177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6702572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6704352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6705937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6707378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6708833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6710281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6711741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6713161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6714615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6716101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6717579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6719012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6720466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6721891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6723417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6724935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6726357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6727837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6729277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6730715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6732138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6733596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6735042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6736482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6737936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6739396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6740832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6742343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6744153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6745611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6747057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6748515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6749963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6751399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6752867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6754299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6755750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6757184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6758657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6760079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6761635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6763147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6764603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6766030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6767464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6768924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6770361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6771806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6773239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6774686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6776128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6777575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6778998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6780585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6782139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6783583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6785007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6786471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6787924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6789347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6790792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6792225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6793713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6795141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6796582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6798033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6799574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6801080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6802514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6804181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6805666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6807079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6808534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6809982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6811433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6812874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6814312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6815781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6817224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6818805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6820347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6821813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6823259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6824709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6826140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6827584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6829038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6830478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6831898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6833336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6834795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6836213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6837731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6839266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6840720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6842146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6843660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6845096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6846606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6848038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6849496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6850928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6852391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6853826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6855255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6856784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6858319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6859763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6861198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6862653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6864085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6865525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6866953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6868417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6869852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6871292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6872724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6874174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6875691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6877197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6878660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6880095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6881559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6882979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6884416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6885855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6887303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6888742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6890188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6891628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6893114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6894614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6896134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6897559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6899018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6900453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6901867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6903316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6905036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6906494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6907914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6909384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6910824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6912256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6913777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6915338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6916778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6918207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6919674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6921112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6922578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6924009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6925457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6926894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6928379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6929819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6931271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6932776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6934322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6935740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6937183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6938638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6940075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6941516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6942947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6944395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6945828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6947273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6948730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6950176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6951679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6953199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6954620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6956074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6957517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6958974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6960407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6961841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6963296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6964710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6966156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6967585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6969058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6970546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6972057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6973485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6974935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6976357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6977796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6979250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6980708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6982119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6983533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6984985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6986428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6987859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6989370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6990901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6992336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6993819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6995253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6996701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6998140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.6999599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7001022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7002460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7003892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7005580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7007008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7008561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7010135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7011551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7012993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7014425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7015869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7017286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7018736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7020168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7021619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7023040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7024465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7025909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7027413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7028953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7030379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7031834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7033281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7034719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7036147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7037592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7039717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7041208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7042633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7044084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7045523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7047038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7048557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7050006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7051460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7052881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7054315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7055747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7057200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7058612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7060067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7061499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7062954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7064368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7065922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7091111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7092732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7094257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7095689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7097109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7098530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7099941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7101341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7102755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7104733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7106192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7107590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7109201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7110746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7112150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7113560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7115004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7116430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7117841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7119262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7120684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7122112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7123535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7124962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7126382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7127894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7129388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7130803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7132224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7133662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7135061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7136478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7137899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7139320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7141070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7142519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7143967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7145440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7146953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7148447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7149870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7151664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7153126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7154557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7155993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7157424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7158831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7160245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7161671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7163110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7164541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7166076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7167579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7169001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7170401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7171817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7173228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7174677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7176088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7177499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7178911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7180334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7181748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7183149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7184671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7186187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7187598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7189012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7190449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7191872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7193330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7194766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7196190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7197611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7199027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7200451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7201861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7203357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7205159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7206595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7208012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7209449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7210853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7212270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7213695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7215121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7216518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7217934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7219356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7220775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7222296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7223825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7225245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7226674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7228095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7229504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7230931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7232356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7233785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7235189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7236604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7238017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7239418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7240905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7242403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7243843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7245246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7246657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7248063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7249492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7250922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7252327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7253763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7255204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7256617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7258035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7259534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7261051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7262461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7263903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7265343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7266769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7268192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7269615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7271036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7272448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7273873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7275289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7276706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7278201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7279693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7281097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7282522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7283966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7285369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7286786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7288212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7289639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7291042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7292460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7293935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7295374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7296854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7298374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7299792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7301223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7302650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7304349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7305792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 11%] 2024-08-06T21:15:45.7307221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7308625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7310026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7311461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7312916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7314332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7315851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7317376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7318800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7320217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7321631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7323072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7324498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7325912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7327311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7328728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7330160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7331570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7333010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7334504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7336013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7337415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7338833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7340248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7341662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7343080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7344490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7345899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7347312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7348722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7350134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7351548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7353062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7354552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7355954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7357383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7358814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7360226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7361639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7363098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7364525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7365936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7367355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7368785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7370217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7371753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7373288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7374715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7376170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7377585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7379015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7380444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7381901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7383319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7384747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7386176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7387633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7389047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7390560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7392091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7393560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7395004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7396423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7397867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7399303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7400740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7402171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7403611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7405331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7406781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7408197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7409726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7411266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7412686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7414121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7415548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7416986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7418391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7419821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7421247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7422710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7424123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7425561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7426980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7428499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7429992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7431401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7432872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7434310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7435741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7437155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7438604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7440045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7441473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7442913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7444357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7445826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7447345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7448836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7450275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7451706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7453139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7454565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7455983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7457429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7458839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7460267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7461689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7463159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7464567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7466075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7467584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7469029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7470450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7471879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7473708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7475162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7476601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7478030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7479468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7480892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7482315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7484291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7485882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7487404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7488829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7490247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7491692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7493175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7494586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7496020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7497450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7498887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7500297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7501733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7503164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7505079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7506623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7508052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7509474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7510923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7512336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7513785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7515204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7516625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7518039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7519448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7520889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7522308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7523839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7525333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7526769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7528186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7529615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7531025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7532462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7533918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7535340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7536766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7538193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7539647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7541061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7542566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7544093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7545547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7546959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7548396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7549823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7551262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7552669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7554121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7555537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7556966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7558400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7559807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7561316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7562823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7564270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7565683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7567215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7568661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7570099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7571533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7572997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7574456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7575908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7577343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7578783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7580319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7581845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7583291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7584732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7586197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7587617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7589050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7590487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7591948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7593433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7594889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7596335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7597785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7599284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7600791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7602242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7603699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7605436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7606878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7608334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7609783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7611228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7612658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7614137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7615589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7617030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7618580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7620137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7621567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7622984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7624454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7625891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7627342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7628767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7630212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7631642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7633097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7634542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7635983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7637484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7639018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7640431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7641853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7643305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7644762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7646263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7647692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7649143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7650575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7652017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7653443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7654893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7656756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7658356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7659769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7661214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7662647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7664068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7665498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7666923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7668371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7669782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7671212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7672650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7674117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7675610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7677129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7678566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7680092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7681517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7682954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7684408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7685858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7687304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7688729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7690175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7691613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7693075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7694593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7696119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7697553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7698987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7700417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7701857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7703282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7705018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7706470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7707906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7709361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7710787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7712222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7713776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7715354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7716773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7718213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7719661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7721116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7722539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7723984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7725435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7726882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7728299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7729723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7731167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7733082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7734624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7736048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7737494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7738937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7740379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7741799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7743245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7744700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7746186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7747615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7749067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7750499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7751996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7753515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7754966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7756424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7757848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7759284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7760696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7762141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7763615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7765062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7766488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7767942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7769351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7770851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7772354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7773778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7775230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7776652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7778102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7779539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7780977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7782410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7783859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7785324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7786766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7788190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7789717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7791239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7792659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7794138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7795594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7797033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7798450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7799892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7801320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7802766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7804460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7805942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7807371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7808938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7810466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7811905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7813336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7814792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7816229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7818183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7819678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7821114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7822549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7823971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7825447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7826881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7828404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7829905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7831345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7832773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7834201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7835629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7837058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7838507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7839917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7841343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7842767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7844220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7845651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7847159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7848668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7850116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7851540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7852983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7854413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7855909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7857336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7858757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7860200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7861643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7863082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7864500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7866032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7867553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7868969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7870384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7871836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7873262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7874699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7876140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7877579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7879000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7880413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7881852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7883278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7884795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7886304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7887728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7889160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7890612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7892022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7893497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7894933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7896399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7897816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7899238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7900674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7902098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7903598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7905376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7906834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7908269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7909700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7911113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7912559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7913997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7915444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7916856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7918296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7919730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7921141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7922693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7924226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7925697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7927118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7928558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7929982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7931436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7932856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7934284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7935718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7937164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7938572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7939998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7941492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7943010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7944427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7945870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7947312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7948734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7950156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7951575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7953020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7954451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7955903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7957329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7958766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7960268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7961831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7963256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7964686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7966158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 12%] 2024-08-06T21:15:45.7967571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7969008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7970432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7971869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7973270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7974693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7976134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7977581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7979068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7980567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7981992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7983426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7984859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7986278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7987717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7989154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7990591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7992008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7993489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7994947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7996383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7997878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.7999408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8000842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8002284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8003713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8005408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8006867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8008284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8009711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8011137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8012590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8014012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8015466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8017038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8018613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8020022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8021459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8022890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8024340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8025778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8027205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8028660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8030094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8031535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8032964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8034413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8035944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8037463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8038881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8040327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8041758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8043174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8044591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8046100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8047540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8048951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8050387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8051822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8053259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8054744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8056273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8057693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8059136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8060547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8061973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8063395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8064848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8066276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8067706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8069138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8070577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8071998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8073487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8075008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8076425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8077845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8079257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8080686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8082105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8083529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8084957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8086387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8087811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8089224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8090644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8092157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8093728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8095174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8096612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8098043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8099487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8100901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8102339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8103787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8105569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8107009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8108461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8109884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8111424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8112959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8114374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8115847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8117289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8118725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8120143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8121591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8123030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8124458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8125899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8127348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8128781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8130285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8131786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8133220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8134677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8136125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8137566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8138997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8140452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8141877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8143319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8144744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8146212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8147617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8149121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8150644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8152098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8153514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8154939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8156393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8157822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8159257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8160679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8162120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8163546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8165040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8166455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8167968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8169483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8170916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8172336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8173766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8175224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8176639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8178076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8179497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8180928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8182334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8183764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8185203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8186728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8188215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8189641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8191063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8192508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8193950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8195391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8196845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8198288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8199722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8201146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8202606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8204300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8205908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8207441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8208889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8210331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8211784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8213207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8214641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8216093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8217508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8218935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8220365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8221818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8223232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8224737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8226281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8227729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8229149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8230591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8232016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8233456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8234868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8236303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8237745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8239177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8240614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8242029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8243544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8245054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8246498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8247919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8249362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8250776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8252195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8253607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8255050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8256475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8257887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8259322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8260743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8262301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8263798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8265234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8266661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8268111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8269519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8270951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8272380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8273832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8275260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8276704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8278138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8279568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8281080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8282579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8284014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8285457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8286887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8288291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8289726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8291153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8292586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8294031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8295497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8296932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8298353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8299857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8301364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8302802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8304472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8305956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8307395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8308842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8310263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8311704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8313129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8314611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8316050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8317483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8319012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8320536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8321953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8323358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8324797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8326248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8327671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8329089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8330528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8331947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8333372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8334787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8336236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8337731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8339240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8340643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8342066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8343518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8344936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8346419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8347854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8349301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8350707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8352142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8353562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8354996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8356483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8357986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8359407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8360829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8362252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8363667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8365109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8366540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8367970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8369377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8370823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8372256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8373689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8375195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8376719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8378149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8379576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8380998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8382425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8383873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8385306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8386744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8388156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8389599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8391009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8392430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8393883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8395427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8396921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8398344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8399767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8401215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8402617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8404287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8405784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8407210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8408642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8410056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8411494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8412919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8414452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8415992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8417437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8418870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8420305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8421724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8423142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8424584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8426009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8427442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8428865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8430301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8431698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8433199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8434710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8436173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8437584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8439016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8440436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8441884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8443301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8444715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8446182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8447618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8449048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8450461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8451970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8453478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8454905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8456347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8457783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8459198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8460620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8462031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8463463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8464881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8466305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8467735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8469152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8470662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8472152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8473576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8474999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8476453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8477854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8479278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8480703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8482135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8483542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8484959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8486407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8487829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8489325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8490823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8492250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8493702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8495124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8496521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8497949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8499368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8500787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8502184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8503615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8505317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8506742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8508274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8509824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8511265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8512684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8514118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8515551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8516994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8518408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8519849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8521272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8522721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8524142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8525570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8527077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8528576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8529997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8531405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8532846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8534268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8535710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8537126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8538568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8539983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8541411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8542824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8544264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8545787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8547277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8548701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8550128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8551579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8552987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8554415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8555863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8557314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8558722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8560159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8561580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8563008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8564515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8566029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8567463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8568882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8570307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8571727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8573160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8574586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8576029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8577436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8578869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8580291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8581705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8583194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8584712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8586155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8587556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8588984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8590407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8591844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8593292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8594729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8596152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8597582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8598985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8600397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8601881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8603394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8605054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8606493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8607940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8609368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8610785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8612191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8613637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8615062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8616511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8617930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8619372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8620797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8622343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8623871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 13%] 2024-08-06T21:15:45.8625312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8626762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8628181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8629614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8631034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8632477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8633880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8635312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8636753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8638196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8639593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8641094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8642595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8644032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8645482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8646916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8648348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8649769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8651194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8652597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8654034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8655476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8656907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8658315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8659837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8661337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8662757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8664168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8665601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8667013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8668415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8669833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8671240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8672668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8674070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8675510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8676923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8678436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8680432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8681851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8683268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8684715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8686129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8687543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8688981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8690399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8691816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8693263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8694715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8696159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8697658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8699158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8700583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8701998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8703417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8705098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8706572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8707999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8709405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8710830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8712252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8713682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8715077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8716629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8718171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8719605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8721013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8722445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8723865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8725303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8726737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8728156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8729588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8731017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8732448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8733856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8735356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8736871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8738284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8739680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8741114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8742536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8743947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8745356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8746813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8748231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8749635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8751065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8752475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8753982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8755468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8756913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8758324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8759760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8761165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8762595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8764012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8765453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8766874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8768288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8769715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8771121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8772529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8774000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8775504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8776941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8778361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8779767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8781192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8782608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8784027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8785424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8786882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8788304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8789702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8791122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8792619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8794192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8795606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8797059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8798482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8799921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8801329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8802761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8804431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8805883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8807298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8808700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8810123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8811652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8813178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8814580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8816014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8817860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8819307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8820715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8822172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8823607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8825039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8826469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8827943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8829382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8830898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8832426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8833856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8835311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8836739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8838177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8839600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8841050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8842459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8843889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8845315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8846779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8848198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8849706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8851258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8852693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8854126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8855548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8857004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8858444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8859887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8861304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8862752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8864193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8865638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8867079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8868619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8870141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8871563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8873011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8874440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8875892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8877331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8878774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8880201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8881648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8883066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8884509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8885932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8887481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8888974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8890414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8891846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8893317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8894746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8896163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8897638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8899066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8900499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8901919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8903372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8905071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8906644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8908197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8909635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8911055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8912482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8913885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8915306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8916750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8918171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8919596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8921014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8922465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8923872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8925371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8926881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8928342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8929763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8931235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8932666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8934122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8935550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8936979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8938431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8939871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8941320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8942743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8944267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8945823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8947271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8948689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8950136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8951566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8953004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8954425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8955849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8957315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8958734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8960177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8961604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8963129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8964634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8966069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8967528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8968990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8970412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8971851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8973286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8974734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8976151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8977603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8979049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8980472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8981977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8983477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8984918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8986352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8987808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8989226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8990674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8992111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8993578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8994998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8996442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8997958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.8999370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9000874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9002405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9003855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9005531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9006982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9008424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9009875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9011299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9012730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9014142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9015585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9016995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9018431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9019952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9021490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9022916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9024332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9025776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9027205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9028631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9030046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9031498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9032931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9034369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9035794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9037256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9038768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9040268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9041710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9043149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9044611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9046029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9047495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9048929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9050372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9051778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9053216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9054648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9056096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9057605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9059129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9060556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9061993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9063432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9064850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9066288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9067738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9069178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9070593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9072040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9073482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9074916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9076402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9077965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9079394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9080832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9082254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9083672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9085110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9086519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9087973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9089390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9090833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9092248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9093714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9095207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9096733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9098154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9099587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9101017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9102465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9103872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9105550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9107006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9108448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9109884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9111310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9112764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9114310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9115858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9117294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9118732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9120160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9121580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9122987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9124415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9125865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9127280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9128723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9130151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9131592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9133072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9134584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9136004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9137471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9138885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9140311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9141735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9143188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9144607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9146027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9147499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9148943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9150380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9151881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9153446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9154866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9156294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9157732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9159166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9160587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9162019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9163439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9164874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9166299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9167737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9169162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9170653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9172177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9173585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9175024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9176450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9177916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9179327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9180759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9182181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9183621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9185033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9186455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9187899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9189389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9190885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9192289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9193760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9195193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9196620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9198040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9199474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9200900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9202322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9203734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9205462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9206892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9208423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9209958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9211383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9212830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9214248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9215690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9217115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9218580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9220024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9221459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9222880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9224322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9225723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9227255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9228793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9230218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9231641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9233058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9234503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9235924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9237362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9238786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9242787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9246160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9247598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9249033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9250587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9252126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9253563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9254988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9256443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9257886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9259321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9260746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9262188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9263630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9265131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9266627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9268077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9269572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9271037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9272497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9273923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9275372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9276783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9278216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9279645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9281101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9282535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9284025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9285539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 14%] 2024-08-06T21:15:45.9286975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9288413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9289881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9291396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9292852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9294338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9295773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9297207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9298637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9300062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9301470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9302983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9304814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9306251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9307692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9309209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9310723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9312134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9313595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9315017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9316453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9317863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9319292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9320711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9322222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9323724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9325138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9326578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9328051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9329526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9330946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9332386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9333818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9335240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9336650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9338080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9339495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9340962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9342438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9343875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9345294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9346748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9348224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9349653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9351108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9352534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9353968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9355401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9356860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9358273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9359762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9361252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9362722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9364142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9365638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9367110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9368536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9369968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9371380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9372835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9374269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9375706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9377118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9378600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9380085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9381513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9382948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9384435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9385914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9387327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9388776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9390210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9391662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9393106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9394595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9396023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9397519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9399004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9400446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9401869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9403399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9405120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9406561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9407992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9409424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9410851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9412266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9413730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9415154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9416653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9418141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9419582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9421003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9422484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9423974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9425416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9426848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9428273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9429680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9431104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9432555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9433977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9435448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9436929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9438366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9439766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9441229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9442698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9444149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9445555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9446987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9448400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9449823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9451254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9452661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9454148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9455683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9457118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9458530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9460021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9461505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9462945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9464368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9465818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9467252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9468684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9470112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9471538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9473039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9474505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9475941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9477366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9478860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9480319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9481746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9483190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9484639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9486047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9487476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9488902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9490329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9491794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9493278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9494761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9496188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9497660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9499123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9500561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9501984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9503440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9505137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9506586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9508008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9509427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9510917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9512409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9513875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9515283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9516768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9518250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9519688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9521097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9522536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9523974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9525416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9526823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9528251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9529716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9531211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9532621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9534052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9535533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9537015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9538446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9539855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9541289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9542708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9544142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9545596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9547037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9548494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9549970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9551378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9552792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9554286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9555744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9557165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9558582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9560028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9561436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9562867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9564309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9565760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9567229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9568732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9570159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9571601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9573054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9574553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9575979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9577395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9578813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9580218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9581650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9583089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9584517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9585919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9587393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9588862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9590281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9591725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9593198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9594676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9596080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9597512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9598937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9600382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9601795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9603240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9604920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9606444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9607930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9609357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9610764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9612259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9613728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9615137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9616586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9618009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9619428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9620833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9622260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9623698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9625161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9626620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9628055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9629477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9630942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9632404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9633869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9635299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9636729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9638138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9639564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9641013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9642418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9644509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9646018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9647457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9648855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9650330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9651794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9653233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9654664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9656096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9657507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9658928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9660359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9661761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9663242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9664747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9666183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9667597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9669083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9670561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9671991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9673410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9674875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9676300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9677731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9679147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9680558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9682035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9683483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9684928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9686344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9687829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9689285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9690713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9692134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9693618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9695046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9696474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9697903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9699329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9700794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9702256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9703702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9705436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9706952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9708434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9709873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9711306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9712747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9714158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9715601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9717029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9718451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9719913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9721414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9722858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9724268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9725753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9727223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9728660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9730071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9731503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9732915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9734360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9735767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9737195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9738653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9740126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9741555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9742966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9744450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9745926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9747350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9748763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9750198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9751613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9753024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9754440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9755874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9757373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9758844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9760253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9761663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9763137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9764609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9766032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9767455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9768907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9770321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9771754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9773184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9774651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9776108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9777594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9779019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9780446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9781917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9783381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9784832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9786257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9787682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9789089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9790526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9791956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9793385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9794851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9796349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9797831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9799256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9800665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9802131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9803625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9805487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9806942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9808369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9809819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9811242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9812675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9814221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9816130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9817645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9819078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9820499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9822006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9823475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9824881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9826344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9827772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9829197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9830613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9832051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9833471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9834939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9836428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9837863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9839282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9840746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9842204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9843630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9845076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9846554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9847993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9849414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9850862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9852269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9853735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9855199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9856655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9858049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9859507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9860987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9862429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9863838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9865248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9866691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9868110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9869536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9870938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9872433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9873920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9875356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9876785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9878274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9879755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9881185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9882608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9884042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9885487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9886902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9888342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9889760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9891239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9892697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9894167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9895619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9897117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9898565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9899994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9901417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9902858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9904659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9906111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9907558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9908983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9910476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9911959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9913404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9914826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9916324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9917798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9919232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9920660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9922090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9923496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9924907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9926363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9927763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9929222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9930686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9932122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9939797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9941547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9943065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9944509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9945917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9947350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9948765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9950193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9951598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 15%] 2024-08-06T21:15:45.9953032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9954500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9955982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9957397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9958819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9960279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9961749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9963194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9964609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9966042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9967458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9968876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9970279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9971710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9973191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9974659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9976063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9977481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9978936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9980381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9981793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9983230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9984671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9986067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9987484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9988907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9990344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9991744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9993234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9994762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9996194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9997597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:45.9999064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0000524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0001940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0003377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0005144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0006603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0008025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0009450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0012135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0014876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0017672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0020325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0022965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0025694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0028455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0031120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0033766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0036454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0039114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0041784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0044435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0047084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0049793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0052504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0055168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0057814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0060507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0063181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0065784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0068449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0071108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0073778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0076427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0079073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0081715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0084416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0087106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0089770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0092458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0095226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0097960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0100627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0103338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0106319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0108977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0111664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0114330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0116988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0119713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0122450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0125117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0127767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0130530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0133229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0135898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0138557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0141196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0143834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0146546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0149206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0151853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0154585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0157364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0160057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0162734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0165453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0168207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0170888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0173570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0176289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0178993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0181667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0184381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0187068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0189799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0192526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0195228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0197906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0200634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0203416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0206337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0209033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0211742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0214419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0217076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0219763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0222493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0225249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0227995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0230717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0233402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0236157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0238908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0241610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0244312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0247006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0249673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0252347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0255040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0257727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0260449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0263177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0265870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0268531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0271240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0273987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0276704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0279407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0282089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0284745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0287433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0290117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0292786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0295552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0298314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0300985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0303674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0306721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0309486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0312172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0314844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0317528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0320221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0322902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0325560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0328217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0331001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0333744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0336417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0339086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0341814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0344519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0347173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0349882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0352599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0355299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0357988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0360672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0363390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0366130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0368870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0371591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0374302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0377047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0379781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0382463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0385154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0387834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0390525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0393212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0395955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0398648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0401363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0404405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0407132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0409833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0412574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0415334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0418033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0420726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0423427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0426132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0428851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0431567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0434255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0437015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0439796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0442491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0445155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0447962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0450736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0453408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0456078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0458770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0461473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0464139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0466843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0469529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0472269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0475006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0477712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0480372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0483104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0485852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0488521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0491204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0493950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0496639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0499317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0502023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0505036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0507815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0510557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0513237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0515901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0518621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0521357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0524046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0526735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0529416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0532067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0534744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0537437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0540125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0542824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0545568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0548263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0550983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0553706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0556451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0559180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0561882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0564557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0567256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0569979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0572678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0575363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0578145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0580884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0583545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0586211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0588937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0591691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0594403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0597122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0599802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0602503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0605480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0608162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0610847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0613599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0616334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0619003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0621687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0624469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0627226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0629900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0632571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0635291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0637974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0640645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0643320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0646001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0648705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0651419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0654108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0656785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0659549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0662260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0664926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0667590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0670271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0672935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0675605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0678323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0681003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0683704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0686439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0689141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0691826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0694587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0697340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0700027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0702712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0705708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0708412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0711096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0713767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0716423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0719077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0721825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0724585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0727264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0729995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0732754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0735407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0738069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0740744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0743439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0746185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0748863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0751552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0754247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0756977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0759712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0762402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0765167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0767889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0770563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0773241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0775928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0778630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0781285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0783986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0786660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0789329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0792036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0794789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0797491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0800166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0802878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0805922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0808620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0811303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0814427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0817162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0819847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0822554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0825243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0828017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0830779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0833468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0836117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0838859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0841625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0844281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0846931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0849640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0852341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0854988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0857647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0860341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0863076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0865795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0868451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0871101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0873833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0876552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0879246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0881935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0884631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0887294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0889972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0892661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0895409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0898176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0900979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0903633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0906649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0909390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0912109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0914795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0917492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0920163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0922799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0925473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0928159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0930830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0933589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0936325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0938988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0941660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0944384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0947112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0949807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0952520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0955175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0957852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0960571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0963252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0965926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0968648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0971381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0974051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0976709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0979433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0982162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0984831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0987504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0990178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0992856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0995572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.0998240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1000909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1003635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1006676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1009346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1012025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1014800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1017549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1020225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1022900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1025610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1028289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1030960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1033665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1036341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1039050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1041773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1044464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1047199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1049921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1052628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1055280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1057964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1060665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1063346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1066018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1068704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1071369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1074044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1076779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1079480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1082182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1084879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1087612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1090270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1092945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1095642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1098295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1100982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1103634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1106582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1109314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1112057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1114738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1117385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1120793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1123559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1126208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1128853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 16%] 2024-08-06T21:15:46.1131538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1134233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1136903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1139589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1142271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1145004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1147738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1150412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1153097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1155839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1158568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1161261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1163935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1166622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1169281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1171936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1174594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1177295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1180009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1182722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1185402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1188108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1190803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1193503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1196228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1198914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1201591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1204537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1207236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1209922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1212608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1215380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1218184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1220879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1223560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1226271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1229001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1231691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1234364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1237026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1239694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1242378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1245054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1247711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1250443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1253174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1255841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1258480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1261202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1263923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1266582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1269276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1271954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1274625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1277279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1279947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1282619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1285342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1288105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1290764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1293415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1296171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1298872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1301509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1304453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1307172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1309820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1312472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1315160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1317841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1320566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1323284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1325957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1328644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1331373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1334119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1336793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1339486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1342192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1344866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1347592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1350287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1352969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1355683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1358407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1361073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1363741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1366425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1367910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1369332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1370735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1372167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1373589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1375043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1376454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1377879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1379334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1380821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1382225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1383652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1385121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1386612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1388016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1389444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1390863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1392288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1393731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1395191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1396621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1398028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1399481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1400941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1402369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1403835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1405555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1406962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1408394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1409817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1411224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1412645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1414076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1415520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1416920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1418425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1419920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1421357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1422765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1424280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1425764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1427200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1428611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1430044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1431446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1432864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1434300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1435699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1437175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1438644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1440062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1441467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1442945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1444429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1445847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1447253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1448689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1450105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1451510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1452943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1454385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1455884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1457343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1458770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1460186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1461662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1463110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1464565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1465980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1467411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1468802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1470220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1471638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1473053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1474532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1475987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1477420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1478834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1480296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1481748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1483179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1484626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1486048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1487448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1488887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1490314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1491717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1493180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1494715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1496162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1497565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1499036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1500487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1501910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1503302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1504996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1506415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1507848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1509252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1510672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1512142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1513626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1515059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1516459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1517946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1519430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1520849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1522255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1523700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1525132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1526552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1527966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1529402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1530864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1532336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1533748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1535161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1536632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1538076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1539494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1540911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1542351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1543774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1545191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1546606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1548042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1549485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1550954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1552389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1553832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1555316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1556785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1558235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1559678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1561122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1562546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1564012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1565452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1566895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1568372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1569864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1571293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1572720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1574199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1575679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1577125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1578541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1579988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1581414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1582864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1584309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1585751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1587274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1588780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1590204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1591637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1593110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1594674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1596096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1597520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1598980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1600417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1601862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1603288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1604996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1606498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1608000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1609422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1610865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1612364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1613858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1615287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1616714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1618168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1619581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1621013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1622441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1623888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1625340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1626823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1628255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1629699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1631154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1632656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1634089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1635547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1636971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1638385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1639821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1641248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1642678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1644136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1645664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1647099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1648520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1649972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1651464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1652884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1654329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1655749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1657197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1658632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1660059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1661500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1662972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1664496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1665928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1667372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1668846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1670351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1671771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1673213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1674665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1676107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1677520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1678955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1680387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1681855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1683341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1684791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1686235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1687704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1689191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1690606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1692060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1693500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1694990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1696412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1697867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1699311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1700777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1702270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1703708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1705465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1706974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1708488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1709911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1711356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1712771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1714203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1715629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1717081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1718497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1719991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1721487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1722925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1724359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1725818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1727308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1728737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1730167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1731588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1733030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1734476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1735916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1737334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1738813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1740292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1741729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1743142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1744617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1746108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1747510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1748935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1750359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1751804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1753214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1754670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1756092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1757580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1759040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1760474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1761901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1763396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1764890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1766315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1767767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1769213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1770655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1772078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1773534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1774998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1776473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1777961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1779405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1780828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1782300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1783768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1785213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1786663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1788086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1789524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1790955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1792411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1793847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1795380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1796859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1798318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1799726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1801204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1802683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1804383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1805829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1807258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1808710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1810144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1811582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1813006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1814528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1816020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1817446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1818850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1820407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1822229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1823665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1825092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1826540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1827964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1829375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1830810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1832232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1833728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1835209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1836652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1838076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1839564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1841027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1842464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1843889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1845365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1846785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1848215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1849635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1851057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1852531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1853988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1855452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1856874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1858338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1859799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1861231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1862655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1864083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1865517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1866956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1868390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1869805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1871230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1872696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1874194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1875630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1877102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1878630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1880076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1881487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1882928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1884349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1885806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1887215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1888646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1890065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1891530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1893006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1894465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1895959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 17%] 2024-08-06T21:15:46.1897437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1898857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1900263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1901707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1903134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1904856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1906320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1907766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1909187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1910686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1912172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1913594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1915044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1916514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1918021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1919437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1920879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1922285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1923712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1925145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1926587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1927995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1929467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1930937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1932358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1933784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1935252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1936744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1938178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1939613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1941037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1942479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1943920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1945372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1946836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1948329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1949814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1951248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1952668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1954130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1955641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1957050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1958485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1959909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1961352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1962766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1964202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1965645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1967133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1968597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1970025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1971450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1972922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1974393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1975837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1977280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1978710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1980137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1981559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1983009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1984441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1985937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1987402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1988835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1990253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1991725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1993183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1994645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1996113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1997524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.1998953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2000371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2001819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2003224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2004999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2006529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2007981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2009394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2010882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2012372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2013826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2015248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2016683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2018125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2019568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2021007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2022421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2023908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2025385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2026813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2028221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2029696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2031166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2032594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2034011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2035457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2036882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2038294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2039730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2041144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2042618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2044073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2045517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2046943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2048427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2049879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2051310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2052733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2054181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2055602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2057022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2058458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2059865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2061325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2062795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2064229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2065670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2067138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2068588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2070012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2071430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2072855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2074260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2075725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2077160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2078571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2080039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2081517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2082966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2084385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2085886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2087362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2088809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2090227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2091666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2093080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2094559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2095995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2097406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2098882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2100358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2101784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2103196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2104969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2106500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2107926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2109344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2110791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2112209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2113637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2115053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2116512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2117996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2119472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2120907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2122336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2123826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2125291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2126736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2128156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2129599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2130998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2132425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2133852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2135290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2136750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2138232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2139655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2141079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2142545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2144010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2145445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2146873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2148299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2149703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2151142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2152573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2153997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2155457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2156945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2158370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2159774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2161241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2162702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2164132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2165549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2166973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2168385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2169818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2171222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2172646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2174097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2175600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2177011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2178434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2179940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2181419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2182845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2184257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2185725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2187163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2188601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2190022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2191466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2193565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2195120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2196572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2198003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2199469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2200956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2202362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2203782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2205533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2206975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2208406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2209833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2211283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2212757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2214252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2215674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2217110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2218592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2220081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2221498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2222920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2224351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2225766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2227202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2228649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2230075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2231529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2233010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2234424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2235838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2237280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2238782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2240198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2241617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2243032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2244449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2245924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2247337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2248771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2250182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2251663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2253114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2254540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2255961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2257444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2258910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2260344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2261767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2263212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2264624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2266040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2267489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2268903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2270355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2271809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2273241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2274659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2276131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2277593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2279030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2280454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2281873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2283278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2284697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2286136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2287549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2289020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2290492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2291930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2293336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2294871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2296353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2297816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2299239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2300668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2302077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2303512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2305209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2306637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2308167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2309660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2311080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2312487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2313971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2315448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2316874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2318306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2319738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2321150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2322566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2323974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2325387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2326888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2328375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2329807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2331223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2332702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2334151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2335575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2336983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2338432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2339820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2341230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2342643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2344068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2345504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2346971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2348424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2349841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2351301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2352753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2354190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2355613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2357042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2358470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2359904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2361327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2362755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2364207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2365679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2367123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2368547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2370021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2371480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2372905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2374299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2375725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2377134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2378585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2379992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2381420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2382874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2384359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2385761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2387169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2388658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2390121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2391535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2392944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2394429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2395859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2397277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2398699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2400135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2401600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2403073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2404747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2406183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2407680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2409155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2410567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2411978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2413413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2414811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2416230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2417643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2419081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2420547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2422034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2423452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2424881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2426325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2427795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2429235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2430657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2432086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2433501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2434937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2436364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2437793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2439200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2440675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2442139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2443554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2444953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2446410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2447902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2449307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2450739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2452156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2453592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2454995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2456413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2457829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2459304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2460749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2462161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2463575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2465050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2466493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2467915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2469353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2470767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2472182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2473589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2475013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2476416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2477879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2479381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2480809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2482220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2483673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2485127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2486539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2487994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2489392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2490800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2492213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2493657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2495093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2496559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2498049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2499495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2500899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2502366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2503829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2505545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2506958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2508397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2509824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2511237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2512656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2514056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2515551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2517038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2518486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2519891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2521372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2522848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2524264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2525670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2527091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2528551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2529952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2531376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2532794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2534275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2535733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2537164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2538604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2540089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2541549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2542977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2544389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2545874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2547279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2548705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2550147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2551562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2553019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 18%] 2024-08-06T21:15:46.2554474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2555920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2557333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2558824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2560282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2561705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2563119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2564545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2565945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2567356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2568818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2570218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2571637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2573090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2574579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2575973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2577434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2578907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2580333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2581722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2583137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2584545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2585972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2587367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2588786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2590204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2591655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2593117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2594553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2595997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2597457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2598949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2600356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2601798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2603220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2604904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2606337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2607765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2609221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2610727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2612268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2622347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2623959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2625519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2627027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2628465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2629894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2631286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2632692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2634098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2635532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2636927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2638398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2639872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2641294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2642686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2644127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2645591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2646999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2648440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2649844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2651254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2652674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2654087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2655484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2656929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2658399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2659807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2661199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2662647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2664122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2665516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2666929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2668351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2669773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2671164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2672580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2673990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2675891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2677380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2678813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2680221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2681681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2683147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2684550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2685980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2687409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2688825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2690228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2691650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2693060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2694575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2696020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2697462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2698873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2700320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2701764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2703161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2705081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2706513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2707950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2709356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2710787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2712183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2713586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2715106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2717174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2718685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2720111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2721614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2723100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2724517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2725925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2727346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2728763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2730174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2731567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2732985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2734434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2735895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2737290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2738717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2740175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2741630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2743028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2744436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2745865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2747255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2748679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2750131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2751563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2753007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2754474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2755888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2757329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2758782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2760251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2761650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2763063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2764485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2765872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2767287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2768710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2770123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2771557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2773029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2774438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2775847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2777287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2778781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2780192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2781593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2783015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2784422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2785854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2787255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2788690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2790194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2791683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2793082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2794552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2796006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2797482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2798883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2800298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2801708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2803123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2804857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2806275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2807705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2809202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2810684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2812088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2813528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2815028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2816527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2818143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2819834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2821280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2822696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2824130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2825560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2827010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2828422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2829942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2831416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2832848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2834287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2835764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2837178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2838617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2840053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2841488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2842913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2844345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2845817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2847235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2848716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2850219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2851647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2853056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2854556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2856032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2857463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2858884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2860352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2861786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2863225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2864647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2866063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2867546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2869002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2870459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2871878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2873358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2874811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2876238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2877662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2879105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2880528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2881968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2883391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2885028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2886777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2888254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2889688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2891406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2892915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2894414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2895861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2897295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2898718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2900149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2901580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2902993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2904700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2906220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2907715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2909146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2910552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2912036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2913506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2914942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2916345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2917767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2919183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2920641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2922059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2923488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2924955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2926450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2927874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2929291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2930797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2932277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2933711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2935133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2936563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2937976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2939401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2940832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2942265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2943731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2945204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2946613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2948026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2949505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2950979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2952410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2953837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2955280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2956681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2958112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2959539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2961004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2962460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2963943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2965371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2966821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2968281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2969754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2971213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2972640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2974070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2975478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2976916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2978349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2979769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2981237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2982729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2984160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2985588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2987046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2988517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2989960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2991388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2992818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2994234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2995724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2997141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.2998568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3000044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3001549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3002961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3004655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3006161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3007663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3009056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3010463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3011901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3013316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3014738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3016150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3017576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3019048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3020532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3021946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3023386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3024857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3026330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3027741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3029188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3030639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3032056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3033493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3034925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3036378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3037830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3039319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3040756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3042193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3043636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3045106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3046520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3047964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3049380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3050811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3052248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3053674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3055099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3056550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3058031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3059454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3060896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3062347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3063848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3065266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3066694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3068110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3069544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3070989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3072398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3073827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3075278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3076750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3078148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3079567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3081043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3082577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3083972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3085399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3086819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3088314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3089711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3091155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3092572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3094030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3095538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3096952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3098387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3099852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3101348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3102759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3104473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3105919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3107347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3108759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3110192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3111624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3113087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3114570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3115987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3117418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3118874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3120363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3121786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3123216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3124621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3126046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3127480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3128924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3130333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3131810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3133277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3134704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3136127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3137593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3139081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3140526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3141957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3143371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3144799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3146257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3147682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3149085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3150521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3151983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3153442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3154864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3157016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3158534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3159934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3161374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3162858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3164293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3165697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3167124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3168544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3169978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3171439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3172920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3174334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3175749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3177210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3178662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3180090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3181522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3182937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3184332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3185760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3187170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3188580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3190017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3191513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3192924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3194322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3195832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3197296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3198733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3200140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3201594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3203430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3205169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3206605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3208035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3209550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3211085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3212522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3213957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3215433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3216905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3218322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3219731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 19%] 2024-08-06T21:15:46.3221169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3222610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3224035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3225450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3226878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3228335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3229800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3231211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3232670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3234137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3235586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3237016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3238449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3239893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3241300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3242753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3244178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3245617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3247070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3248549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3249962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3251395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3252853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3254323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3255738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3257163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3258594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3260022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3261461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3262903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3264319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3265762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3267244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3268669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3270095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3271548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3273053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3274477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3275906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3277325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3278748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3280199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3281614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3283050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3284513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3286001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3287399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3288822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3290277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3291768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3293170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3294634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3296061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3297518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3298939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3300351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3301789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3303251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3304958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3306394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3307824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3309307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3310803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3312215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3313654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3315078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3316502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3317911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3319324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3320753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3322212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3323693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3325107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3326534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3327965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3329432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3330840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3332296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3333698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3335126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3336547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3337972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3339399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3340848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3342363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3343798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3345225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3346680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3348186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3349610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3351037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3352476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3353909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3355328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3356755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3358162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3359574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3361059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3362544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3363969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3365383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3366858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3368311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3369737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3371163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3372630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3374036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3375474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3376898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3378326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3379795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3381261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3382722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3384150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3385669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3387608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3389075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3390500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3391919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3393339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3394817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3396244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3397664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3399134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3400607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3402040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3403474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3405337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3406841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3408276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3409682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3411107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3412518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3413971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3415378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3416799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3418285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3419789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3421191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3422599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3424094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3425556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3426964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3428363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3429785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3431191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3432602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3434021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3435446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3436895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3438353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3439757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3441175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3442657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3444108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3445535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3447011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3448462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3449875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3451304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3452732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3454178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3455629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3457111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3458524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3459955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3461391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3462848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3464284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3465704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3467124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3468523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3469953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3471377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3472817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3474267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3475746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3477167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3478584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3480027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3481486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3482938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3484344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3485778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3487187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3488620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3490029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3491454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3492918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3494405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3495838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3497252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3498703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3500184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3501576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3503002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3504690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3506115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3507530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3508940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3510374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3511856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3513354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3514763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3516193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3517685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3519179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3520584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3522003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3523469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3524882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3526313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3527753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3529184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3530617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3532079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3533511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3534935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3536368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3537837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3539241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3540674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3542073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3543500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3544925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3546341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3547756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3549152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3550639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3552111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3553552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3554957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3556425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3557884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3559308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3560715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3562201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3563652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3565046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3566468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3567873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3569364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3570812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3572228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3573653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3575123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3576569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3577987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3579394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3580824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3582212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3583631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3585068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3586476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3587927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3589380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3590810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3592223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3593694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3595266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3596689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3598098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3599508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3600894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3602298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3603744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3605366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3606859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3608337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3609772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3611165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3612632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3614121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3615551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3616951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3618373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3619785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3621204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3622628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3624051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3625521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3626987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3628407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3629810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3631289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3632761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3634180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3635578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3637015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3638428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3639839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3641241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3642652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3644124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3645570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3646993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3648404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3649881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3651332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3652766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3654186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3655632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3657034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3658457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3659876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3661302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3662768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3664227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3665654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3667059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3668508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3669960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3671386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3672813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3674235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3675641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3677065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3678483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3679895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3681297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3682761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3684309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3685710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3687174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3688645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3690086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3691491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3693037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3694774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3696235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3697651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3699081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3700485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3701961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3703448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3705725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3707204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3708739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3710231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3711635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3713069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3714512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3715931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3717340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3718764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3720170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3721639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3723106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3724536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3725972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3727410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3728875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3730286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3731724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3733121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3734558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3735958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3737386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3738773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3740221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3741681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3743098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3744531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3746025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3747504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3748912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3750325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3751732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3753158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3754601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3756030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3757432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3758911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3760384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3761810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3763217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3764702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3766192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3767596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3769022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3770439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3771871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3773264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3774711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3776123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3777595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3779058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3780484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3781897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3783351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3784830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3786540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3788006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3789433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3790851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3792255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3793699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3795179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3796773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3798565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3800023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3801439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3802920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3804878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3806341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3807779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3809178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3810606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3812022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3813463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3814862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3816393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3817879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3819316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3821168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3822697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3824183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3825596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3827026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3828434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3829866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3831280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3832703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3834113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3835539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3837018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3838491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3839891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3841309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3842759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3844194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3845610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3847037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3848465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3850178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3851618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3853024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3854447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3855918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3857413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3858827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3860262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3861701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3863171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3864590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3866021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3867458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3868864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3870299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3871729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3873150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3874599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3876123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3877547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 20%] 2024-08-06T21:15:46.3878961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3880399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3881880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3883299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3884702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3886128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3887977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3889455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3890861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3892277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3893744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3895274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3896695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3898113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3899570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3901053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3902449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3903867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3905575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3906999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3908416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3909826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3911243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3912727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3914205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3915598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3917026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3918491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3919978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3921373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3922800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3924217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3925607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3927035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3928443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3929868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3931303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3932773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3934180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3935602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3937054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3938524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3939924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3941352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3942755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3944175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3945579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3947003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3948419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3949851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3951325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3952741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3954148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3955587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3957075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3958484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3959890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3961289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3962720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3964135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3965533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3966973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3968383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3969860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3971308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3972731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3974145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3975624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3977090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3978516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3979920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3981343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3982727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3984126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3985555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3986985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3988491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3989949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3991374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3992782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3994244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3995732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3997178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.3998591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4000000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4001389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4002817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4004501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4005909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4007421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4008901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4010339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4011740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4013233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4014694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4016108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4017530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4018944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4020403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4021830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4023222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4024616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4026077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4027560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4028969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4030350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4031819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4033281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4034692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4036092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4037602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4039016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4040430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4041842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4043258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4044733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4046234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4047687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4049084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4050554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4051996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4053408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4054810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4056290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4057709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4059125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4060534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4061964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4063352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4064797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4066303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4067750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4069188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4070647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4072142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4073578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4075019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4076438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4077905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4079339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4080775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4082191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4083650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4085142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4086541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4087993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4089460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4090953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4092366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4093801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4095257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4096708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4098141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4099575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4100994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4102486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4104207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4105660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4107108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4108638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4110137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4111559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4113010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4114444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4115874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4117291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4118743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4120159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4122262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4123773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4125264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4126687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4128140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4129632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4131055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4132501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4133913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4135340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4136760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4138220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4139636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4141115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4142587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4144038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4145445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4146904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4148388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4149821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4151258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4152675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4154099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4155516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4156935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4158347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4159823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4161291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4162713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4164120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4165591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4167054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4168466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4169895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4171326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4172770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4174181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4175623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4177051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4178551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4180016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4181457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4182882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4184374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4185840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4187251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4188708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4190136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4191557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4192968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4194411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4195873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4197344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4198820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4200326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4201754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4203222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4204940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4206405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4207840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4209280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4210713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4212139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4213850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4215279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4216804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4218300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4219770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4221181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4222672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4224149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4225590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4226997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4228410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4229864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4231286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4232714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4234129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4235609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4237089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4238516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4239944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4241425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4242890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4244309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4245718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4247161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4248586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4250013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4251442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4252865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4254344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4255802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4257233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4258638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4260125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4261576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4262991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4264403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4265844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4267248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4268652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4270102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4271526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4272948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4274398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4275892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4277315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4278742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4280227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4281764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4283199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4284630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4286045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4287487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4288924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4290355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4291789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4293253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4294777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4296178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4297609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4299078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4300583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4301988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4303420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4305293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4306771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4308183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4309626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4311043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4312542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4314038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4315444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4316883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4318854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4320574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4321996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4323441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4324868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4326289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4327702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4329138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4330558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4332010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4333475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4334891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4336324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4337763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4339229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4340653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4342093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4343497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4345016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4346483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4347931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4349335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4350819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4352304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4353732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4355156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4356605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4358086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4359505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4360952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4362371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4363788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4365200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4366622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4368017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4369488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4370972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4372378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4373800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4375256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4376802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4378202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4379627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4381069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4382506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4383909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4385341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4386761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4388343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4390097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4391570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4392988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4394471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4395985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4397400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4398830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4400257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4401696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4403101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4404869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4406311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4407811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4409295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4410730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4412171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4413631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4415116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4416527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4417961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4419360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4420779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4422216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4423656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4425058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4426540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4428161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4429605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4431002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4432472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4433942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4435346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4436755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4438154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4439579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4440992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4442431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4443832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4445250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4446701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4448233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4449628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4451059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4452541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4453991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4455412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4456837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4458276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4459679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4461109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4462551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4463991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4465440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4466920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4468325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4469752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4471193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4472657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4474088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4475512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4476937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4478338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4479768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4481209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4482637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4484078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4485559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4486980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4488396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4489846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4491347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4492784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4494195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4495663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4497078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4498522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4499989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4501413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4502874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4504592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4506002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4507418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4508893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4510385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4511779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4513200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4514632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4516045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4517459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4518867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4520297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4521763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4523252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4524663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4526100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4527561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4529034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4530500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4531952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4533384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4534794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 21%] 2024-08-06T21:15:46.4536220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4537634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4539067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4540503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4541977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4543392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4544820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4546257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4547729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4549133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4550560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4551971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4553374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4554798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4556215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4557626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4559064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4560538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4561969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4563382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4564836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4566314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4567725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4569138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4570550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4571990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4573404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4574801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4576210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4577611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4579090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4580530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4581968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4583376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4584886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4586324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4587738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4589155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4590596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4592017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4593425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4594898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4596316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4597783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4599245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4600676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4602114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4603580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4605282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4606722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4608140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4609554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4610949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4612403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4613823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4615220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4616710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4618187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4619622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4621015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4622526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4623999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4625435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4626838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4628263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4629682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4631121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4632555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4633965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4635437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4636924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4638355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4639769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4641255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4642742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4644154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4645557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4647037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4648452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4649866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4651275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4652722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4654181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4655632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4657057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4658465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4659936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4661378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4662817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4664235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4665668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4667059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4668480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4669895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4671323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4672781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4674235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4675658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4677059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4678502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4679945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4681368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4682797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4684208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4685602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4687018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4688429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4689836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4691392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4692875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4694309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4695740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4697208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4698679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4700111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4701515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4702969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4705138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4706623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4708037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4709465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4710868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4712402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4713891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4715845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4717395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4718886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4720301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4721703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4723139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4724572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4725994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4727399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4728828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4730238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4731696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4733144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4734562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4735993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4737451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4738935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4740347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4741788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4743185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4744612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4746014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4747442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4748827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4750279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4751737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4753165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4754570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4756006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4757472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4758879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4760289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4761685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4763104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4764528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4765945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4767344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4768820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4770287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4771708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4773111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4774576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4776069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4777465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4778885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4780290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4781708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4783094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4784513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4785918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4787385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4788826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4790245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4791652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4793131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4794576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4796010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4797442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4798860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4800273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4801668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4803112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4804996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4806825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4808322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4809762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4811171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4812641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4814118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4815518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4817315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4818737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4820153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4821559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4822995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4824395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4825865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4827320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4828748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4830135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4831588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4833038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4834498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4835934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4837329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4838755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4840163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4841578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4842977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4844404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4845866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4847331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4848722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4850137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4851953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4853431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4854828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4856237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4857661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4859048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4860455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4861857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4863274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4864706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4866165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4867575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4869005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4870441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4871900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4873312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4874746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4876167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4877570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4878997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4880474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4881887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4883326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4884839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4886249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4887650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4889082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4890908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4892343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4893752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4895228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4896643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4898074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4899472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4900889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4902376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4903879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4905769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4907216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4908731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4910257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4911667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4913098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4914532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4915979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4917411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4918835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4920276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4921751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4923253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4924667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4926124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4927589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4929067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4930476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4931914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4933342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4934745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4936195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4937621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4939068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4940518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4942001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4943429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4944872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4946395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4947891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4949366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4950810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4952229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4953665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4955079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4956526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4957949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4959393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4960878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4962303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4963722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4965174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4966682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4968104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4969525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4970937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4972375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4973795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4975217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4976645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4978059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4979539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4980992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4982424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4983883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4985373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4986794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4988543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4989986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4991420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4992817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4994237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4995686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4997118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.4998914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5000521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5001948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5003366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5005300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5006800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5008256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5009688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5011114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5012527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5013972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5015406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5016834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5018393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5020111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5021773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5024544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5027290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5029996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5032736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5035417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5038071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5040748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5043441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5046103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5048769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5051516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5054248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5056976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5059639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5062362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5065080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5067761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5070446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5073132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5075830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5078514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5081174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5083874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5086610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5089370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5092120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5094796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5097541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5100271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5102931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5105920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5108639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5111338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5113991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5116669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5119356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5122105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5124844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5127543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5130244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5133616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5136368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5139066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5141757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5144435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5147112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5149780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5152473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5155141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5157864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5160578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5163250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5165926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5168618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5171328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5173996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5176664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5179316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5181953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5184661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5187328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5189973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5192713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5195496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5198167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5200837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5203594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5206656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5209343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5212032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5214730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5217407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5220137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5223227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5225921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5228690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5231433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5234078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5236752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5239578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5242349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5245010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5247756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5250428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5253105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5255777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5258449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5261149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5263858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5266565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5269242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5271925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5274663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5277380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5280065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5282749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5285397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5288064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5290718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5293406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5296098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5298744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5301460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5304902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5307587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5310327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5313062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5315724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5318361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5321016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5323675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5326351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5329036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5331685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5334344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5337134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5346285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5349003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5351696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 22%] 2024-08-06T21:15:46.5354501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5357227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5359880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5362534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5365202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5367844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5370483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5373131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5375810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5378503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5381228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5383872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5386523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5389212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5392359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5395362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5398045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5400695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5403354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5406520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5409205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5411843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5414598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5417358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5420046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5422702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5425406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5428116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5430774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5433416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5436061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5438705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5441366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5444018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5446635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5449321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5452019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5454672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5457335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5460024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5462707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5465357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5467994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5470651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5473317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5475948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5478575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5481234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5483968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5486658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5489299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5491974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5494649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5497360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5499973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5502604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5505674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5508295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5510943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5513592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5516241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5519002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5522296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5525137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5527866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5530670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5533405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5536073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5538743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5541416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5545103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5548709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5551484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5554158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5556893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5559638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5562306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5566062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5568826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5571564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5574211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5576860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5579513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5582207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5585739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5589192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5591880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5594523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5597365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5600123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5602776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5606789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5609648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5612397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5615052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5617729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5620412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5623077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5626579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5629277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5632797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5637069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5639888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5642560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5646365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5649120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5651838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5655335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5659196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5662784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5666448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5670738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5674196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5677974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5682494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5687544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5692179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5695958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5699102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5702283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5706534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5710024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5713178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5716459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5719535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5722853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5726527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5730287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5733951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5737187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5739826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5742561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5745320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5747971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5750625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5753288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5755919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5758569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5761209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5763850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5766557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5769250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5771926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5774563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5777281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5779986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5782621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5785265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5787917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5790527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5793147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5795839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5798525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5801205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5803886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5806982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5809639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5812384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5815090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5818228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5820954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5823604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5826299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5828970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5831652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5834338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5837092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5839840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5842529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5845203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5847972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5850686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5853747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5856403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5859052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5861712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5864389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5867048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5869692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5872405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5875115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5877769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5880428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5883144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5885868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5888532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5891164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5894306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5897046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5899710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5902383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5905534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5908233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5911004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5913743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5916412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5919078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5921803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5924532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5927236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5929902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5932560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5935228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5937889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5940556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5943221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5945904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5948625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5951293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5953937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5956619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5959324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5962019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5964688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5967335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5970017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5972691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5975328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5977998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5980685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5983387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5986026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5988669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5991340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5994043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5996784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.5999426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6002067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6005159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6007824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6010469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6013118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6015895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6018624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6021277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6023949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6026694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6029416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6032064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6034764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6037456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6040104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6042765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6045433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6048095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6050808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6053518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6056173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6058824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6061529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6064553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6067229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6069907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6072590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6075209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6077889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6080561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6083234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6085948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6088654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6091331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6093977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6096791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6099503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6102173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6105582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6108255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6110930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6113596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6116236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6118876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6121610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6124320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6126963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6129611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6132319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6135042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6138185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6140835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6143507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6146340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6149107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6151756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6154458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6157119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6159839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6162533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6165184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6167872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6170564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6173246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6175872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6178531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6181175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6183794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6186477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6189135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6191757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6194439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6197170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6199837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6202485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6205682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6208421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6211090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6213765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6216408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6219393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6222117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6224772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6227411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6230676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6233497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6237198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6242443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6247954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6251477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6255234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6258470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6262137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6265824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6269344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6273084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6277327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6281247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6285454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6289047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6293315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6296858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6301356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6305766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6308834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6311857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6314500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6317128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6319755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6322431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6325217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6327935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6330584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6333222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6335892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6338577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6341199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6343843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6346476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6349129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6351733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6354370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6357015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6359622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6362318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6365021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6367630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6370324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6373020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6375672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6378316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6380965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6383635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6386275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6388923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6391556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6394613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6397403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6400092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6402685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6405834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6408589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6411314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6413940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6416571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6419201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6421834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6424459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6427104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6429749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6432447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6435123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6437790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6440447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6443130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6444581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6446047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6447481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6448903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6450304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6451721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6453128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6454813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6456437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6458783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6460268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6462139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6464081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6465574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6466997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6468389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6469809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6471207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6472634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6474026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6475437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6476884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6478381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6479781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6481201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6482704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6484239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6485683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6487081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6488503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6489910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6491312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6492705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6494126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6495624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6497079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6498473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6499885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6501327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6502762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6504643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6506080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6507502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6508881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6510286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6511689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6513112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6514491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6515998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6517475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6518891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6520274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6521716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6523184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6524574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6525975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6527358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6528766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6530163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6531561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6532953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6534400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6535846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6537247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6538625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6540087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6541537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6542934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6544341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6545751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6547177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 23%] 2024-08-06T21:15:46.6548568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6549985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6551389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6552880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6554320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6555741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6557136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6558595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6560025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6561417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6562851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6564264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6565667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6567051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6568471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6569876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6571313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6572758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6574181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6575582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6577027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6578466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6579871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6581305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6582721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6584145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6585554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6586985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6588380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6589794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6591233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6592712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6594092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6595545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6596992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6598449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6599852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6601248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6602682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6604562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6606015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6607409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6608825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6610313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6611787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6613186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6614607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6616070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6617556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6618948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6620352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6621775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6623170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6624582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6625980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6627389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6628807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6630275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6631670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6633099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6634523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6636012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6637401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6638799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6640206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6641588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6643022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6644433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6645843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6647276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6648757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6650175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6651580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6653057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6654545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6655950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6657342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6658767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6660168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6661590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6662993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6664403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6665842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6667323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6668712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6670114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6671547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6673038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6674420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6675813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6677232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6678635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6680037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6681432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6682870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6684269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6685805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6687265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6688679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6690083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6691545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6693001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6694414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6695858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6697249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6698648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6700051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6701479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6702874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6704789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6706294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6707721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6709103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6710564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6712026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6713461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6714851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6716253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6717670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6719074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6720490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6721885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6723361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6725165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6726604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6728001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6729469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6730912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6732302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6733700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6735382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6737125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6738530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6739941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6741338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6743591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6745080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6746668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6748070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6749537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6750970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6752376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6753774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6755202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6756614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6758007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6759427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6760835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6762245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6763679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6765175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6766567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6767963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6769383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6770844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6772242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6773645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6775054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6776442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6777862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6779250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6780647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6782081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6783590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6784991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6786393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6787830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6789295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6790676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6792086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6793486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6794907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6796344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6798020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6799453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6800905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6802347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6803729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6805572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6807079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6809459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6812373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6814161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6815585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6816968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6818396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6819860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6821282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6822664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6824196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6825684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6827107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6828525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6829985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6831435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6832861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6834278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6835677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6837091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6838518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6839920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6841300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6842773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6844228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6845624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6847012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6848493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6849944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6851446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6853076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6854882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6856345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6857752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6859211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6860631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6862141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6863600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6865030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6866447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6867933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6869399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6870821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6872230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6873667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6875061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6876466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6877906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6879343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6880809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6882262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6883700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6885111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6886566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6888013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6889462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6890885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6892313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6893714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6895603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6897084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6898487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6899998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6901466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6902906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6904950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6906513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6908005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6909429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6910842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6912261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6913665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6915091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6916488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6917889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6919437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6920976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6922391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6923785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6925246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6926704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6928111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6929504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6930955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6932365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6933775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6935178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6936586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6938050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6939496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6940932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6942323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6943782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6945216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6946613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6948011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6949437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6950841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6952249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6953656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6955079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6956463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6957899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6959377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6960804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6962227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6963674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6965155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6966575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6968002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6969409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6970853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6972277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6973697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6975100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6976543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6978023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6979414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6980832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6982294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6983777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6985174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6986596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6988006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6989449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6990843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6992299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6993709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6995180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6996669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6998078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.6999514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7000980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7002444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7003853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7005836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7007260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7008681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7010097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7011524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7012927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7014553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7016041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7017449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7018878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7020337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7021815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7023221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7024657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7026058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7027475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7028877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7030327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7031718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7033172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7034633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7036063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7037455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7038892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7040378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7041789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7043203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7044905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7046397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7047810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7049222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7050632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7052111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7053586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7054996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7056379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7057820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7059290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7060701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7062111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7063525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7064961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7066357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7067775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7069188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7070686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7072138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7073565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7074972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7076428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7077896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7079302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7080748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7082160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7083570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7084961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7086391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7087805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7089214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7090726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7092207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7093617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7095030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7096512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7097963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7099388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7100806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7102224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7103635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7105535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7106953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7108371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7109861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7111380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7112776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7114194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7115647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7117118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7118522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7119917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7121362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7122769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7124182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7125580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7127001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7128462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7129922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7131339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7132763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7134211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7135663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7137061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7138473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7139904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7141316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7142742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7144160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7145589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7147026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7148501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7149900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7151342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7152767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7154219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7155618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7157028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7158443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7159831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7161268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7162680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7164091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7165525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7166999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7168408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7169815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7171303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7172876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7174283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7175683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7177109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7178514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7179946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7181368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7182789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7184228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7185696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7187089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7188494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7189936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7191447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7192836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7194246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7195686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7197104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7198508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7199909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7201348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7202748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7204764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 24%] 2024-08-06T21:15:46.7206255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7207677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7209089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7210557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7212034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7213459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7214874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7216272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7217690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7219093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7220514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7221918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7223362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7224812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7226239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7227629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7229080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7230524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7231966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7233354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7234744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7236376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7237818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7239236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7240631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7242130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7243593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7245036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7246505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7247973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7249431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7250842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7252262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7253683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7255083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7256470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7257887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7259286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7260749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7262206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7263626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7265029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7266512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7267949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7269362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7270774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7272226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7273616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7275019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7276456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7277874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7279331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7280783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7282229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7283645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7285101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7286552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7287968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7289373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7290784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7292188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7293612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7295029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7296459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7297920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7299376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7300805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7302206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7303661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7305574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7307022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7308419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7309841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7311253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7312697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7314099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7315507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7316933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7318427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7319915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7321317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7322741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7324210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7325675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7327075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7328505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7329919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7331336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7332746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7334154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7335580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7337019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7338529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7339933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7341376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7342816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7344283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7345695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7347191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7348601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7350015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7352160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7353885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7355326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7357546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7359061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7360475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7361882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7363328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7364796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7366199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7367609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7369011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7370425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7371829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7373256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7374646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7376116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7377585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7379430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7381405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7383012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7384567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7385968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7387404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7388822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7390256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7391654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7393079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7394488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7396474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7397959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7399359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7400779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7402231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7403694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7405560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7407002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7409137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7410724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7412513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7414298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7415734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7417312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7418804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7420243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7421657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7423124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7424617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7426029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7427478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7428885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7430316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7431719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7433153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7434543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7435948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7438393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7439964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7441366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7442768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7444239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7445701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7447115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7448535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7449958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7451362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7452773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7454173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7455599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7457052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7458528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7459931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7461354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7463629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7465191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7467273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7468727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7470173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7471553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7472964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7474371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7475791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7477254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7478719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7480143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7481726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7483952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7485467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7486895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7488320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7489747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7491165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7492598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7494023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7495448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7496958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7498445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7499870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7501416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7503568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7505872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7507958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7509433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7510856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7512257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7513709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7515110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7516525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7518054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7519567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7521034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7522450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7523936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7525791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7527796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7529238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7530670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7532082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7533490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7534913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7536341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7537746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7539330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7540782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7542193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7543593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7545059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7546494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7547897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7549328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7551488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7553023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7555044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7556535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7557922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7559418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7560878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7562300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7563686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7565155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7566627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7568044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7569803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7571792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7573257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7574676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7576093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7577507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7579009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7580467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7581872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7583259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7584719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7586167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7587591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7589205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7591166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7593254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7594769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7596241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7597654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7599189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7600641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7602055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7603461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7605454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7606960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7608497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7610788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7612276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7613691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7615093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7616515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7617923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7619448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7620944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7622373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7623781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7625237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7626680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7628098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7629504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7630923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7632338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7633738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7635205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7637331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7638790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7640943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7642541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7643928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7645335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7646845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7648323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7649715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7651107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7652538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7653927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7655328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7656718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7658128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7659564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7661019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7662472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7664760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7666305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7667760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7669145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7670569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7671978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7673384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7674798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7676206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7677633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7679063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7680524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7681931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7683376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7684812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7686273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7687660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7689076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7690509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7692604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7694105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7696240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7697696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7699167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7700672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7702077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7703482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7705530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7707065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7708473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7709879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7711278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7713393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7715109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7716544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7717972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7719512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7721191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7722595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7724007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7726128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7727657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7729390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7730830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7732236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7733644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7735050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7736449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7738182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7740658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7742915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7744450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7745875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7747285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7748740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7750202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7751631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7753048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7754468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7755867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7757276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7759581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7761096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7762610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7764071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7765498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7766886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7768335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7769779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7771223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7772619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7774038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7775434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7776837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7778245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7779639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7781187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7782698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7784801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7786249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7788458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7789986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7791393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7792806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7794238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7795644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7797086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7798505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7799959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7801727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7803207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7805147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7807311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7809117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7810590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7812341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7813835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7815282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7816668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7818067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7819478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7820887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7822292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7824308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7825870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7828090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7829586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7831756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7833288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7834699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7836121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7837540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7838955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7840357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7841746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7843148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7844589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7846243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7848371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7849848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7851339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7852827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7854215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7856016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7857460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7858889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7860284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7861687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7863107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7864577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7866402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7868265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7870409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7872001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7873478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7874897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7876317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7877729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7879157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7880547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7881953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7883374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7885145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7886633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7888041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7889469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7891883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7893426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7894828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 25%] 2024-08-06T21:15:46.7897107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7898560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7899967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7901383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7902811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7904631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7906172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7907672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7909084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7910491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7912021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7913511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7914902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7916306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7917694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7919342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7921454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7922917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7925025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7926479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7927988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7929442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7930845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7932249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7933718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7935154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7936565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7937975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7939406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7940794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7942209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7943620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7945032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7947518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7949083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7950509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7951912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7953361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7954798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7956213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7957621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7959031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7960421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7961838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7963247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7964636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7966093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7967549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7968970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7970350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7971800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7974020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7975511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7977626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7979106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7980502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7981924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7983315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7984726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7986199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7987650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7989049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7990431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7992015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7994250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7995705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7997146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7998574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.7999974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8001369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8002761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8004641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8006078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8007594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8009100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8010501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8011922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8014097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8015715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8017849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8019346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8020755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8022166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8023557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8024978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8026357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8027835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8029350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8030765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8032168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8033598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8035061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8037367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8038909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8040299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8041728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8043136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8044547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8045927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8047409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8048893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8050285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8051708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8053158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8054630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8056015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8057427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8058825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8060236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8062068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8063819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8065934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8067504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8068954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8070348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8071755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8073210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8074661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8076050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8077460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8078865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8080446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8082559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8084057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8085461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8086857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8088332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8089792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8091206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8092591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8094062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8095509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8096975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8098362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8099766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8101873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8103383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8105846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8107320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8108835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8110315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8111718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8113103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8114586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8116075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8117476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8118869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8120300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8122609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8124087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8125488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8126935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8128429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8129917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8131335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8132729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8134186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8135612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8137037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8138435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8139860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8141247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8142655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8144056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8145477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8146921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8148997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8150594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8152456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8154264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8155761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8157215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8158679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8160127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8161558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8163012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8164454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8165895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8167360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8168875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8170309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8171716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8174006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8175630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8177086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8178502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8179964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8181397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8182850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8184274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8185714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8187189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8188697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8190121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8191559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8193032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8194518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8196001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8197434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8198904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8200347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8202039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8204408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8206112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8216616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8218292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8219756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8221207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8222717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8224230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8225917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8228225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8229738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8231205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8232631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8234063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8235510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8237017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8238513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8240086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8241866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8243421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8244921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8246405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8247946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8249608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8251055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8252505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8253956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8255364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8256830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8258346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8259779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8261208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8262685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8264177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8265598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8267026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8268449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8269902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8271340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8272800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8274222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8275671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8277151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8278628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8280066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8281542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8283068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8284491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8285937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8287368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8288813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8290221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8291676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8293109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8294560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8296061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8297554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8298979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8300451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8302126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8303550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8305490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8306965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8308407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8309830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8311282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8312744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8314183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8315705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8317233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8318667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8320143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8321644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8323094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8324537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8325954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8327385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8328810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8330264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8331685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8333141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8334624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8336118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8337537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8339013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8340494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8341923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8343374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8344801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8346255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8347690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8349129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8350552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8351988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8353532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8355023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8356441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8357915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8359395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8360814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8362227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8363674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8365123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8366534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8367971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8369396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8370840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8372291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8373789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8375221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8376715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8378191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8379625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8381059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8382519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8383950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8385382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8386837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8388273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8389709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8391179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8392681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8394113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8395609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8397166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8399011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8400491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8401937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8403380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8405244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8406726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8408145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8409582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8411107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8412632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8414063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8415500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8416987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8418498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8419913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8421348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8422778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8424246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8425668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8427093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8428529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8429991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8431463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8432869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8434334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8435803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8437279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8438694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8440137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8441569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8442991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8444431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8445879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8447306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8448762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8450246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8451673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8453121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8454597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8456119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8457743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8459210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8460629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8462058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8463486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8464942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8466344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8467831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8469307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8470734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8472167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8473641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8475144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8476592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8478020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8479437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8480877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8482306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8483747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8485163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8486645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8488130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8489546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8490981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8492455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8493955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8495387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8496869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8498294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8499774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8501183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8502604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8504553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8506118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8507603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8509031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8510450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8511943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8513421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8514862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8516311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8517736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8519155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8520576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8522024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8523442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8524931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8526400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8527835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8529266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8530729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8532188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8533600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8535061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8536463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8537877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8539294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8540734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8542136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8543600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8545106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8546595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8548006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8549481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8550953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8552385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8553816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8555254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8556694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8558130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8559565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8560982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8562463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8563945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8565398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8566821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8568300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8569774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8571198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8572620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8574088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8575539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8576954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8578389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8579809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8581295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 26%] 2024-08-06T21:15:46.8582756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8584183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8585615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8587099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8588555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8589975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8591405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8592846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8594280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8595708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8597198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8598628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8600102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8601579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8603014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8604949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8606480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8607951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8609404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8610831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8612256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8613662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8615104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8616558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8617965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8619464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8620958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8622405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8623821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8625313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8626866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8628311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8629733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8631179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8632606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8634060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8635502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8636920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8638394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8639864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8641289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8642697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8644179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8645722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8647145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8648556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8649998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8651423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8652853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8654270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8655730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8657194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8658653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8660083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8661503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8662997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8664456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8665901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8667386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8668834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8670235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8671667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8673081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8674512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8675976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8677434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8678870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8680288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8681745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8683215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8684646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8686083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8687510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8688909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8690349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8691777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8693206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8694654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8696206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8697649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8699066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8700547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8702023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8703470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8705323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8706787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8708203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8709633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8711039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8712455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8713948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8715456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8716870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8718279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8719772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8721261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8722687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8724102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8725560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8727003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8728513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8729979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8731426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8733586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8735102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8736550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8738000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8739480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8740953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8742385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8743801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8745244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8746676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8748109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8749537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8751293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8752785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8754270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8755691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8757156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8758623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8760107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8761526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8762949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8764377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8765786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8767250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8768679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8770112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8771566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8773056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8774484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8775906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8777376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8778857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8780272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8781669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8783094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8784508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8785948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8787379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8788809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8790221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8791699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8793152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8794574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8796080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8797602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8799007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8800438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8801867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8803299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8805184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8806633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8808100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8809533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8811046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8813005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8814471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8815891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8817410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8818906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8820354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8821790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8823323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8825002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8826439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8827959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8829569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8831077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8832575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8834011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8835414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8836886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8838359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8839804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8841219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8842651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8844070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8845502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8846983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8848443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8849922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8851387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8852804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8854206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8855687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8857586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8859057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8860475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8861910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8863328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8864756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8866173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8867592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8869117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8870581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8872006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8873433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8874919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8876378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8877812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8879265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8880710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8882118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8883550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8884967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8886377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8887833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8889305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8890740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8892160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8893633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8895085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8896561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8897991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8899839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8901284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8902726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8904593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8906043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8907575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8909083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8910555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8911971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8913464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8915012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8916533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8917948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8919381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8920812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8922253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8923662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8925083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8926542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8928031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8929439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8930861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8932336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8933801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8935222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8936632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8938072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8939488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8940919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8942328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8943767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8945240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8946712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8948125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8949544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8951087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8952545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8953973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8955385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8956829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8958225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8959647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8961079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8962513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8963950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8965420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8966834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8968264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8969716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8971193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8972627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8974051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8975485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8976893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8978333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8979771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8981209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8982665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8984155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8985587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8987012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8988466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8989943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8991380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8992785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8994212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8995633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8997110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8998524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.8999972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9001431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9002917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9004752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9006208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9007727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9009262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9010704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9012134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9013599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9015044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9016488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9017924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9019385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9020909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9022431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9023868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9025311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9026787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9028269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9029685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9031162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9032609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9034030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9035474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9036916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9038377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9039837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9041352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9042793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9044250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9045728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9047230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9048666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9050122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9051575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9053009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9054469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9055918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9057368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9058843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9060339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9061792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9063226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9064695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9066193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9067631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9069069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9070525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9071979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9073417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9074847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9076293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9077771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9079272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9080707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9082145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9083620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9085132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9086554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9087995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9089425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9090896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9092321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9093765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9095193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9096702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9098182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9099597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9101058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9102528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9104413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9105910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9107350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9108784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9110220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9111664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9113124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9114572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9116089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9117600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9119048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9120502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9122067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9123580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9125020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9126484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9127918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9129361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9130788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9132263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9133676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9135155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9136644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9138087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9139522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9140994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9142517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9143947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9145388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9146866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9148334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9149777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9151227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9152671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9154168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9155660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9157108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9158536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9160038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9161574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9163002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9164447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9165879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9167332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9168745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9170192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9171635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9173135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9174604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9176046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9177474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9178967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9180440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9181880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9183325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9184767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9186202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9187616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9189077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9190516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9192031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9193515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9194965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9196434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9197915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9199389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9200827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9202277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9203693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9205564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9207007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9208461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9209878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9211301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9212830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9214351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9215763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9217255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9218759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9220218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9221642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9223096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9224550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9225995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9227444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9228875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9230327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9231813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9233322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9234747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9236236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9237739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9239171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9240601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9242058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9243899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9245701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9247168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 27%] 2024-08-06T21:15:46.9248606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9250059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9251548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9253099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9254529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9256601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9258120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9259557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9260987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9262445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9263877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9265321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9266757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9268196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9269636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9271095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9272585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9274028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9275455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9276908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9278396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9279825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9281258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9282680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9284141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9285579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9287009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9288420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9289892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9291388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9292801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9294259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9295733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9297278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9298701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9300142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9301570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9303023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9304904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9306377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9307797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9309306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9310814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9312227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9313671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9315174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9316661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9318080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9319532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9320965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9322398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9323824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9325294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9326722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9328197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9329673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9331105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9332557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9334048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9335563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9336995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9338457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9339882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9341321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9342745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9344191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9345615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9347088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9348563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9350007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9351426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9352884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9354371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9355805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9357243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9358662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9360104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9361540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9363508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9365039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9366556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9368050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9369483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9370903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9372376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9373883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9375326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9376770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9378192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9379630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9381037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9382472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9383889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9385389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9386851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9388287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9389707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9391189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9392646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9394060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9395530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9397013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9398446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9400136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9401757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9403196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9405160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9406729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9408179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9409615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9411108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9412588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9414030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9415457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9416887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9418316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9419740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9421193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9422601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9424071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9425539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9427005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9428416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9429888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9431359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9432806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9434221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9435637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9437098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9438529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9439968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9441378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9442857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9444338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9445770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9447251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9448737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9450208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9451632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9453049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9454491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9455910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9457416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9458857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9460276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9461759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9463221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9464655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9466077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9467572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9469032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9470456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9471886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9473342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9474760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9476209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9477651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9479088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9480573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9482041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9483478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9484902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9486377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9487842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9489282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9490708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9492141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9493551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9494989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9496477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9497898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9499371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9500844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9502287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9503695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9505637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9507165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9508620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9510042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9511489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9512910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9514352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9515768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9517231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9518706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9520192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9521611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9523015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9524489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9525957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9527403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9528816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9530251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9531665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9533088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9534502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9535952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9537444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9538909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9540345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9541771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9543259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9544744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9546183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9547637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9549098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9550510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9551951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9553374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9554816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9556304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9557807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9559234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9560660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9562128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9563594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9565027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9566456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9567908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9569320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9570772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9572203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9573634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9575092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9576590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9578032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9579463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9580926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9582403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9583851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9585267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9586724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9588151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9589593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9591002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9592434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9593897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9595392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9596855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9598293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9599760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9601260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9602666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9604507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9605988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9607437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9608858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9610274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9611713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9613216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9614716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9616130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9617585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9619083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9620582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9621996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9623405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9624849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9626249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9627695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9629222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9630719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9632164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9633645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9635066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9636502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9637971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9639455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9640878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9642333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9643747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9645206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9646657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9648122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9649556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9650974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9652476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9653967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9655404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9656863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9658376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9659796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9661215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9662627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9664048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9665491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9666907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9668360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9669781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9671269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9672732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9674162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9675578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9677091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9678566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9679992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9681416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9682863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9684271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9685691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9687129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9688562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9690027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9691508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9692935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9694348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9696595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9698092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9699527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9700953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9702376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9703781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9705683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9707432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9708867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9710397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9711892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9713331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9714744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9716230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9718029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9719747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9721209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9722641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9724915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9726464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9727889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9729303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9730844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9732327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9733738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9735142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9736621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9738087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9739515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9740947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9742390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9744412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9746221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9747716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9749476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9751032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9752538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9753954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9755381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9756874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9758329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9759799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9761711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9763211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9765384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9766937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9769069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9770565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9772077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9773564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9775005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9776424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9777892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9779344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9780771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9782193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9783712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9785954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9787464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9788892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9790311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9791806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9793305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9794732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9796178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9797695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9799166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9800599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9802317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9804365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9806360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9808400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9809982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9811436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9813418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9814973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9816382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9817808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9819302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9820787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9822205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9823889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9825360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9826784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9828205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9829641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9831657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9833590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9835112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9836533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9837966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9839435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9840947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9842354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9843768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9845209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9846605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9848027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9849450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9850900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9852341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9853815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9855232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9856663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9858106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9860127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9862195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9863672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9865806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9867283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9868726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9870156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9871592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9873096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9874607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9876033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9877456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9878908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9880383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9881804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9883222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9884632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9886054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9888403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9889884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9891309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9892731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9894264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9895728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9897208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9898630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9900116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9901993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9903478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9905352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9906804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9908229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9909645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9911083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9912782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9914925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9917158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9918649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9920069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9921615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9923093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9924528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 28%] 2024-08-06T21:15:46.9925968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9927389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9928788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9930201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9931803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9933959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9935527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9937018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9938458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9939855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9941314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9942768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9944202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9945608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9947054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9948463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9949880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9951304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9953437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9955035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9957156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9958677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9960078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9961590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9963053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9964825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9966500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9967954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9969386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9970868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9972271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9973687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9975190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9976869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9979033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9980534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9982065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9983529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9984952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9986371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9987822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9989228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9990679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9992094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9993510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9994965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9996481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9997919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:46.9999330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0001371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0003067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0005646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0007149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0008573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0009980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0011412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0012857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0014279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0016217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0017782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0019226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0021033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0023185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0024753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0026190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0027597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0029031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0030443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0031872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0033274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0034713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0036120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0037581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0039053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0040456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0042758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0044498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0046438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0047955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0049397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0050814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0052230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0053636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0055071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0056501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0058000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0059470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0060881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0063186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0064770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0066604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0068260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0071013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0073662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0076291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0078941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0081583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0084221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0086951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0089668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0093170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0096646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0099461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0102167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0105349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0108067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0110723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0113398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0116083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0119660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0122474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0125569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0128380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0131069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0133748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0136470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0139202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0141881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0144536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0148023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0151392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0154217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0156872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0159535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0162296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0165054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0168296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0171616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0174386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0177084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0179727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0182396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0185084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0187741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0190411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0193060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0196587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0200131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0202849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0206161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0208827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0211559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0215210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0217971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0220656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0223321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0225960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0228616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0231285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0233934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0236701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0239436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0242911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0246444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0249256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0252037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0254713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0257395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0260047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0262710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0266324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0269240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0271896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0274544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0277287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0279988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0282592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0286091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0289657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0292386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0295049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0297762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0300415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0303821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0307054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0309717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0312385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0315172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0317879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0320534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0323422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0327179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0330354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0333054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0335716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0338372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0341013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0343662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0346346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0349045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0352750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0355469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0358123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0360771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0363466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0366177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0368836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0371465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0374096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0376732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0380177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0383547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0386261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0388999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0391731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0394380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0397070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0400030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0403194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0407282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0409973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0412640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0415288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0418012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0420639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0423267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0426147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0428882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0431512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0435021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0438660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0441377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0444005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0446703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0449372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0452020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0455603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0458283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0460947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0463680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0466412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0469076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0471739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0475261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0478740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0481402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0484076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0486748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0489380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0492110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0495653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0498366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0500999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0503752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0507005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0509657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0512299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0515064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0517829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0520757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0524110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0527608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0530307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0533020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0535705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0538388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0541221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0544904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0547653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0550342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0553122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0555850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0558508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0561215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0564668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0568078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0570784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0573465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0576142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0578905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0581839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0585262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0587990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0590788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0593514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0596199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0598972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0601641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0605514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0608999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0611778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0614469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0617263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0620047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0622718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0625389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0628130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0631019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0634551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0637242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0639907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0642581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0645279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0648052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0650719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0653494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0656217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0658881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0662394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0665921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0668698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0671377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0674037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0676737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0679449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0682152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0685319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0688526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0691281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0694003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0696725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0699406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0702134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0705334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0708010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0710685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0714362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0717682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0720366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0723119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0725904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0728686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0731445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0734641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0737888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0740733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0743498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0746196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0748902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0751594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0754296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0757124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0761028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0764061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0766844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0769587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0772828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0775560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0778376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0782026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0784745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0787466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0790138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0792817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0795509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0798254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0800906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0804408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0808129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0811601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0815025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0817881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0820639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0823322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0825993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0832432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0836979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0839694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0842363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0845082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0848603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0852127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0854897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0857581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0860726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0863513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0866158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0869864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0872581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0875413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0878188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0880856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0883522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0886191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0888849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0891537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0895010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0898601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0901387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0904977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0907706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0910363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0913157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0915906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0918574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0922016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0924961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0927629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0930315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0933026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0935814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0938558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0941254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0943927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0946660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0949851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0953308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0956544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0959240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0961972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0964670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0967357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0970032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0972859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0976475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0979152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0981816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0984485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0987256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0989996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0992652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0996090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.0999563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1002270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1005402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1016245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1019333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1022119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1024795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1027480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1030169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1032924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1035723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1038386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1041930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1045388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1048119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1050784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1053440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1056260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1058941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1061613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1065287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1067968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1070718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1073455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1076113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1078776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1081431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1084891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1088316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1091001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1093750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1096482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1099233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1101916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1106001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1108824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1111573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1114251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1116917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1119583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1122275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1124934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 29%] 2024-08-06T21:15:47.1128430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1131935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1134749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1137420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1140083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1142760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1145505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1148245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1150928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1154558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1157298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1159981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1162644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1165297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1167984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1170763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1173487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1176162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1178891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1182076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1185239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1188646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1191330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1193988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1196623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1199361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1202037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1205191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1208972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1211794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1214483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1217151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1219899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1222631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1225308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1227991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1231417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1234884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1237639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1240333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1242993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1245741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1248555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1251231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1254851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1257648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1260401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1263057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1265720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1268373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1271038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1273701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1276382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1280095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1283243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1286498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1289142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1291812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1295358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1298193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1300880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1303563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1307093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1310663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1313361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1316054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1318957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1321973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1324772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1327427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1330097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1332764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1335503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1338265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1341823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1345261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1347980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1350645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1353349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1356015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1358756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1362166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1365181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1367861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1370517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1373306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1376064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1378719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1381417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1384942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1388405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1391103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1393773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1396522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1399306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1402478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1406297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1408992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1411823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1414579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1417228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1419917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1423282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1426883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1429624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1432328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1435129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1437910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1440590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1443267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1445955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1449638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1452437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1455115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1457809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1460511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1463174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1465842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1468513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1471193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1473879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1476539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1480095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1483555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1486424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1489159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1491839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1494504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1497215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1500008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1503569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1506655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1509458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1512209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1514881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1517574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1520302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1524168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1527375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1528812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1530227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1531652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1533059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1534511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1535910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1537406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1538878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1541171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1542684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1544098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1545652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1547171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1548601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1550031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1551477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1552879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1554306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1555752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1557201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1558649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1560605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1562344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1564491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1566081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1567582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1569018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1570443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1571866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1573274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1574709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1576128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1577569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1579017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1581395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1582888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1584298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1585814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1587314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1588776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1590189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1591629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1593054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1594498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1595906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1597401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1598884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1601144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1603345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1605377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1606951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1608452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1609868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1611287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1612719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1614135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1615560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1616972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1618408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1620900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1622489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1623908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1625343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1626813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1628268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1629689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1631122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1632568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1633975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1635404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1636824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1638264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1640345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1642017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1644140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1645636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1647034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1648528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1650003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1651425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1652850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1654254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1655682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1657101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1658516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1660884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1662490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1663932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1665360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1666777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1668277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1669755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1671169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1672609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1674050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1675499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1676918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1678357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1679810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1682048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1683778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1685673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1687131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1688657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1690117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1691551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1692971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1694394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1695836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1697303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1698741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1701119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1702662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1704474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1705954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1707403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1708929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1710410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1711845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1713271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1714688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1716091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1717527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1718957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1720799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1722723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1724852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1726365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1727761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1729290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1730757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1732192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1733593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1735011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1736420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1737838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1739271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1741592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1743188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1744614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1746038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1747442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1748921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1750409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1751832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1753241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1754659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1756064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1757477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1758875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1760866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1762608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1764769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1766253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1767665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1769184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1770632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1772071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1773489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1774924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1776321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1777743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1779513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1781764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1783347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1784828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1786264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1787685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1789154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1790613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1792036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1793463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1794873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1796271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1797753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1799180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1801372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1803113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1805382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1806868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1808268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1809809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1811300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1812729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1814610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1816073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1817490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1818921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1820783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1822748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1824342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1825889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1827314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1828722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1830144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1831607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1833068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1834463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1835889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1837302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1838709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1840260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1842314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1844352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1845997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1847492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1848914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1850365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1851820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1853288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1854708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1856151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1857565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1858994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1861314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1863185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1864705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1866199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1867607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1869029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1870443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1871895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1873374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1874788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1876209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1877612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1879056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1881282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1882782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1884966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1886513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1887936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1889338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1890779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1892248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1893758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1895169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1896608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1898083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1899703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1901868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1903365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1905213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1907198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1908707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1910129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1911543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1913041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1914527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1915935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1917369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1918783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1920513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1922469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1924604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1926119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1927623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1929092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1930537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1931961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1933412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1934911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1936332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1937778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1939306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1941538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1943012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1944450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1945881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1947389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1948873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1950310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1951716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1953197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1954663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1956111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1957533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1958936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1961128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1962624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1964742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1966199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1967740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1969226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1970651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1972056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1973550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1975017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1976418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1977865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1979489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1981615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1983061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1984483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1985895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1987409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1988893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1990316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1991726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1993158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1994594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1996057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1997530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.1998984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.2001161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.2002967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.2005134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.2006606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.2008156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.2009653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.2011114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.2012533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.2013957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.2015423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 30%] 2024-08-06T21:15:47.2016913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2018325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2020682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2022201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2023617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2025056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2026457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2027966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2029433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2030873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2032295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2033716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2035180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2036667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2038067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2039947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2041725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2043889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2045388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2046807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2048333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2049829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2051255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2052669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2054108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2055566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2057029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2058433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2060759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2062244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2063650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2065090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2066501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2067938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2069421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2070908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2072320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2073753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2075217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2076682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2078098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2080297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2082016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2083895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2085370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2086795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2088225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2089717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2091202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2092617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2094028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2095489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2096968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2099328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2100823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2102226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2103650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2105452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2106902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2108321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2109862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2111381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2112787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2114222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2115703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2117213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2119031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2120889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2123020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2124540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2125947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2127366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2128809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2130329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2131811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2133213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2134647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2136065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2137526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2139872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2141396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2142824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2144245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2145646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2147142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2148558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2150095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2151577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2152994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2154427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2155828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2157306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2159532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2161136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2163157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2164628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2166036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2167461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2168858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2170334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2171834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2173252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2174667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2176066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2178317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2179922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2181338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2182767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2184195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2185606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2187025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2188427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2189905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2191372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2192796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2194225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2195640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2197123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2199344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2201170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2202919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2204782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2206206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2207619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2209023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2210458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2211973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2213456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2214908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2216324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2218299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2220281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2221752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2223178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2224598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2226016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2227446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2228861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2230279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2231758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2233224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2234662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2236083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2237814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2239766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2241861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2243355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2244771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2246179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2247629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2249026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2250441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2251923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2253414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2254808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2256205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2258533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2260098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2261565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2262959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2264394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2265806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2267219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2268645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2270067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2271518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2272980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2274683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2276131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2278329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2279917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2282065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2283522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2284960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2286718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2288151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2289557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2291003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2292472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2293944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2295352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2297062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2299200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2300739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2302249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2303669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2305509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2306931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2308362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2309782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2311196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2312709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2314205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2315614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2317028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2319215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2321160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2322988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2324418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2325843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2327251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2328683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2330081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2331493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2332924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2334462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2335933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2338268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2339778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2341303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2342798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2344225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2345700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2347143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2348579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2350002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2351440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2352865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2354332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2355830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2357946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2359510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2361737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2363273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2364708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2366169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2367600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2369037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2370465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2371915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2373335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2374810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2376476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2378652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2380136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2381669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2383156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2384609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2386031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2387467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2388913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2390339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2391772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2393183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2394667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2396193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2398359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2400300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2402087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2403585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2405429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2407266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2408732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2410196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2411605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2413041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2414466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2416316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2418500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2419988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2421436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2422997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2424481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2425915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2427332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2428769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2430169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2431600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2433042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2434465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2436766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2438780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2440870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2442425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2443965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2445429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2446935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2448382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2449820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2451244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2452698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2454164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2455589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2458029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2459593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2461056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2462476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2463960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2465498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2466942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2468350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2469778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2471207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2472653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2474072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2475509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2477797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2479642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2481483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2482929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2484380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2485919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2487433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2488851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2490302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2491738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2493172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2494598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2496946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2498598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2500082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2501522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2502943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2504801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2506343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2507862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2509288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2510737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2512150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2513580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2515005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2517075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2518968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2521217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2522702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2524128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2525560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2527053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2528550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2530001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2531434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2532849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2534290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2535720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2538055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2539627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2541141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2542560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2543960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2545382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2546839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2548322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2549725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2551178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2552596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2554034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2555441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2557638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2559531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2561381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2562843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2564294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2565725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2567220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2568711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2570134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2571586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2573042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2574479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2576796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2578314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2579818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2581300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2582714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2584181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2585608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2587098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2588564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2589988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2591444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2592861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2594314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2596422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2598057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2600260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2601804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2603234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2605102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2606560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2608092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2609585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2611018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2612457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2613877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2615515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2617727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2619220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2620745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2622261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2623686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2625112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2626522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2628009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2629478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2630906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2632323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2633745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2635316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2637412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2639423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2641138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2642650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2644062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2645501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2646925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2648422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2649886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2651318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2652738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2654191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2656459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2657930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2659389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2660817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2662329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2663819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2665260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2666683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2668155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2669629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2671073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2672505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2673930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2676092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2677586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2679728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2681219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2682745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2684225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2685673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2687083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2688560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2690019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2691478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2692884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2694453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2696714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2698261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2699689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2701102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2702643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2704504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2705963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2707372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2708906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2710396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2711892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2713321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2714762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2717000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2718738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2720619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2722075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2723632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 31%] 2024-08-06T21:15:47.2725147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2726582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2727993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2729480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2730932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2732350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2733761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2736133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2737611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2739035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2740452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2741895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2743388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2744852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2746315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2747802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2749235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2750689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2752185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2753612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2755638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2757286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2759450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2760949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2762384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2763886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2765362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2766802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2768218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2769643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2771107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2772599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2774008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2776341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2777826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2779281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2780692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2782123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2783625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2785131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2786540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2787950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2789395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2790905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2792378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2793792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2795997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2797814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2799675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2801123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2802556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2804435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2805989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2807396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2808809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2810258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2811731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2813245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2815612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2817137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2818542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2819964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2821401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2822847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2824339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2825824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2827245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2828688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2830093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2831572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2833069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2834952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2836715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2838843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2840340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2841767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2843215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2844700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2846191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2847612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2849031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2850439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2851901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2853412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2855715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2857225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2858645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2860078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2861481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2862902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2864787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2866305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2867708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2869131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2870546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2872029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2873475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2875697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2877523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2879302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2880739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2882138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2883561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2884970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2886490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2887957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2889385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2890799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2892273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2894584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2896112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2897636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2899055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2900485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2901908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2903584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2905476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2907058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2909000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2910485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2911899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2913423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2915723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2917246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2919364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2920843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2922284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2923709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2925139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2926547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2928071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2929555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2930983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2932384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2933858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2936225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2937724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2939139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2940577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2942079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2943486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2944925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2946353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2947876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2949338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2950794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2952203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2953673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2955850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2957367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2959579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2961075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2962486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2963893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2965325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2966736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2968236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2969723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2971154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2972579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2974485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2976437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2977912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2979333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2980753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2982161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2983613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2985034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2986442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2987944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2989412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2990842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2992238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2993672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2995890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2997446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.2999620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3001104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3002514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3004324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3005821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3007230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3008801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3010309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3011734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3013144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3014589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3017131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3018783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3020202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3021644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3023073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3024500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3025921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3027377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3028839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3030295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3031718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3033136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3034575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3036841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3038540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3040530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3042020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3043425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3044849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3046266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3047762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3049287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3050754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3052186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3053604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3055738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3057472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3058972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3060414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3061851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3063259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3064687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3066102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3067515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3068953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3070444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3071857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3073254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3074673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3076939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3078862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3080598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3082063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3083482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3084917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3086320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3087744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3089161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3090684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3092164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3093579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3095406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3097467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3099042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3100510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3101944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3103374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3105261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3106689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3108120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3109534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3111045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3112571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3114024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3115493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3117785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3119862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3121516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3122964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3124367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3125806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3127226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3128664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3130066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3131570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3133039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3134488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3136814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3138411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3139896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3141338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3142748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3144145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3145592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3147072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3148487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3149888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3151354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3152821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3154234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3156089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3157944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3160118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3161581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3163030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3164449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3165887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3167296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3168726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3170136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3171648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3173107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3174532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3176904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3178428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3179917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3181384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3182799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3184217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3185630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3187025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3188465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3189877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3191330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3192783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3194215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3196135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3197905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3200155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3201708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3203133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3205007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3206471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3207888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3209326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3210749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3212278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3213750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3215471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3217549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3219024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3220557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3222083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3223484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3224908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3226314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3227728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3229147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3230547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3232051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3233534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3234971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3236939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3238646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3240881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3242434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3243881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3245338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3246776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3248196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3249964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3251411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3252911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3254391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3256095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3258273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3259776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3261356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3262847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3264267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3265736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3267147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3268580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3270007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3271444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3272869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3274327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3275970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3278064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3280065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3281790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3283448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3285081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3286528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3287963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3289400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3290828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3292253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3293667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3295900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3298034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3299506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3300945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3302449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3304329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3305847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3307292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3308732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3310183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3311594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3313024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3314449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3315978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3318295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3319804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3321944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3323558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3325076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3326496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3327937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3329357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3330796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3332204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3333639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3335056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3336683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3338887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3340363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3341830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3343327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3344807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3346233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3347751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3349172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3350659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3352105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3353572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3354990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3356531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3358708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3360685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3362366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3363917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3365421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3366849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3368275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3369684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3371126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3372556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3374011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3375425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3377857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3379418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3380857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3382273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3383744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3385265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3386678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3388118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3389547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3391000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3392412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3393854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3395307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3397353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3399138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3401319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3402818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3404851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3406367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3407800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3409738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3411206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3412634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3414044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3415487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3417073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3419560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3421145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3422584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3424006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3425472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3426950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3428395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3429823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3431239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 32%] 2024-08-06T21:15:47.3432673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3434097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3435546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3437002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3439341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3441287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3443069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3444491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3445921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3447421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3448922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3450343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3451758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3453192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3454612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3456038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3458372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3459984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3461502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3462942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3464354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3465803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3468132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3469661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3471085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3472565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3474008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3475430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3476874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3479084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3480994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3482842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3484327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3485757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3487207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3488686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3490181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3491602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3493052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3494486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3495900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3497953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3499763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3501307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3502791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3504663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3506152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3507593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3509143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3510660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3512090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3513524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3514935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3516386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3518082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3520093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3522239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3523893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3525337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3526746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3528191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3529654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3531144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3532549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3533979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3535396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3536845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3539206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3540704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3542220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3543708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3545140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3546553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3547992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3549482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3550963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3552380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3553813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3555230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3556645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3558658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3560315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3562560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3564085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3565512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3566931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3568370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3569844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3571319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3572741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3574186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3575598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3577027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3579373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3580898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3582404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3583912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3585340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3586769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3588202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3589665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3591233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3592648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3594071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3595477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3596911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3599163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3600924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3602808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3604872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3606399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3607808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3609235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3610718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3612221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3613645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3615080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3616499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3618915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3620409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3621840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3623261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3624812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3626279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3627710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3629123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3630588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3632049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3633453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3634911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3636328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3638100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3639976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3642074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3643611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3645117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3646607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3648127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3649556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3651020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3652506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3653933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3655385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3656828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3659163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3660680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3662132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3663547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3665085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3666564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3668025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3669429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3670900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3672369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3673793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3675217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3676627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3678835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3680346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3682495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3683968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3685501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3686983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3688405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3689831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3691319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3692796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3694228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3695645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3697633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3699512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3700962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3702409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3703824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3705900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3707394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3708843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3710262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3711716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3713185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3714676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3716098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3717531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3719801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3721503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3723441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3724909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3726451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3727940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3729381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3730811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3732248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3733704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3735209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3736641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3738450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3740452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3741913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3743375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3744786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3746307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3747785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3749225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3750660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3752083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3753555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3755091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3756497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3758078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3760118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3762109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3763747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3765180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3766710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3768188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3769614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3771024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3772464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3773931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3775427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3776914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3779184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3780652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3782069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3783473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3784884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3786792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3788288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3789713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3791122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3792559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3794041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3795526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3797439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3799103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3800517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3801942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3803369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3805273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3806881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3808667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3810150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3811584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3813015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3814539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3816039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3817461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3818921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3820956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3822425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3823848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3825271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3826681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3828165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3829684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3831092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3832512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3833977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3835468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3836876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3838305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3839747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3841195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3842612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3844044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3845470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3846957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3848417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3849851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3851283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3852739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3854200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3855605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3857111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3858536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3859974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3861387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3862823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3864243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3866107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3867595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3869010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3870459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3871904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3873380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3874802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3876239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3877642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3879069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3880485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3881925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3883332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3884799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3886275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3887710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3889113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3890558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3892032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3893451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3894865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3896267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3897691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3899159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3900582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3901981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3903460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3905365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3906816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3908269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3909794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3911710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3913153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3914589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3916014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3917456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3918863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3920323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3921737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3923266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3924745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3926169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3927586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3929028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3930493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3931959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3933399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3934833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3936261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3937673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3939117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3940558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3942023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3943476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3944914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3946336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3947827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3949290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3950781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3952226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3953637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3955065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3956471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3957905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3959303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3960798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3962264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3963699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3965103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3966527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3967980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3969461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3970886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3972295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3973728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3975141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3976569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3977972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3979449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3980940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3982371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3983783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3985220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3986690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3988164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3989574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3991011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3992443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3993838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3995260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3996675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3998192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.3999645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4001090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4002501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4004295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4005889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4007378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4008790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4010230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4011654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4013061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4014501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4015923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4017353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4018818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4020337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4021775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4023194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4024644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4026139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4027552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4028969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4030367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4038327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4039999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4041438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4042851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4044391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4045880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4047286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4048723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4050186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4051673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4053070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4054496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4055910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4057345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4058758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4060181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4061594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4063068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4064519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4065930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4067352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4068857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4070316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4071708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4073133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4074541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4075955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4077346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4078782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4080190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4081645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4083095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4084529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4085943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4087348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4088813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4090279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4091717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4093128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4094563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4095977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4097445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4098918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4100391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4101850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4103284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4105314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4106754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4108322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4109817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4111238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4112646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4114074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4115488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4116906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4118318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4119806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4121281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4122701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4124102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 33%] 2024-08-06T21:15:47.4125535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4126999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4128462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4129890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4131313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4132746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4134147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4135576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4136981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4138460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4139901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4141317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4142726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4144155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4145594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4147043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4148488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4149909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4151326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4152724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4154155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4155569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4156982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4158444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4159932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4161348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4162771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4164223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4165762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4167183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4168613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4170039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4171440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4172916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4174308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4175720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4177174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4178692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4180092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4181514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4182966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4184446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4185847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4187250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4188705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4190119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4191525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4192952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4194385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4195840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4197304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4198768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4200203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4201615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4203071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4204938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4206447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4207876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4209289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4210703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4212108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4213538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4215004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4216483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4217886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4219333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4220726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4222189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4223660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4225094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4226501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4227899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4229353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4230776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4232195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4233640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4235116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4236533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4237949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4239378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4240845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4242295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4243703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4245099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4246504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4247983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4249470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4251283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4252807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4254297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4255695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4257114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4258519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4260018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4261462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4262883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4264303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4265723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4267138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4268545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4270000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4271422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4272884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4274332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4275755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4277163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4278612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4280071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4281497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4282913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4284326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4285724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4287133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4288566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4289988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4291448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4292904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4294328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4295716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4297169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4298656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4300113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4301510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4302924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4304859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4306311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4307739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4309148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4310677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4312150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4313559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4314953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4316376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4317840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4319324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4320742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4322165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4323676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4325591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4327017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4328432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4329928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4331400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4332823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4334239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4335721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4337166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4338637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4340051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4341515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4342915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4344336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4345740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4347143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4348587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4350029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4351475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4352889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4354299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4355742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4357216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4358627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4360043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4361465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4362886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4364297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4365685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4367093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4368537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4370060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4371483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4372898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4374343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4375817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4377213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4378632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4380028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4381467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4382853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4384260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4385658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4387107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4388559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4389948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4391390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4392853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4394304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4395692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4397104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4398561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4399966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4401382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4402810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4404756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4406271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4407768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4409180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4410610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4412460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4414014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4415481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4416901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4418292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4419703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4421103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4422557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4423950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4425381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4426854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4428412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4429829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4431223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4432712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4434169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4435575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4436975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4438402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4439816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4441234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4442658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4444105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4445580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4446980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4448399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4449796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4451263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4452717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4454125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4455525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4456948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4458331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4459737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4461132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4462579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4464981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4466493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4467913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4469309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4470760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4472198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4473761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4475174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4476580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4477967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4479387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4480801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4482209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4483661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4485110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4486522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4487899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4489341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4490781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4492195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4493599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4495006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4496404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4497821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4499248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4500658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4502096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4503575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4505431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4506855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4508284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4509792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4511280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4512679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4514172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4515584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4516994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4518396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4519814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4521268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4522732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4524143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4525541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4526961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4528390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4529843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4531242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4532665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4534072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4535477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4536904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4538343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4539800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4541278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4542719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4544179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4545609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4547116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4548628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4550057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4551488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4552903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4554354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4555839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4557255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4558662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4560136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4561611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4563022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4564469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4565925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4567407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4568816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4570250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4571670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4573111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4574534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4575959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4577379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4578866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4580327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4581758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4583245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4584726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4586197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4587611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4589049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4590466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4591885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4593315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4594756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4596177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4597641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4599137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4600580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4601997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4603486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4605409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4606849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4608289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4609683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4611105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4612520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4613975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4615384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4616900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4618392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4619830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4621230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4622712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4624205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4625619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4627034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4628573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4630000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4631415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4632836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4634252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4635716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4637178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4638598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4639998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4641477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4642939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4644388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4645805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4647229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4648673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4650081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4651518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4652939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4654446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4655912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4657345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4658759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4660194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4661633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4663099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4664538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4665966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4667391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4668800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4670239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4671659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4673187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4674676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4676119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4677541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4678966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4680426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4681936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4683373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4684829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4686247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4687670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4689122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4690538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4692018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4693482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4694942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4696362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4697791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4699290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4700777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4702178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4703609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4705468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4706939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4708367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4709778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4711287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4712771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4714198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4715624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4717063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4718542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4720022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4721438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4722875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4724295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4725731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4727142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4728551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4730018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4731465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4732879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4734291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4735736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4737175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4738640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4740045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4741477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4742873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4744294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4745728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4747153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4748629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4750103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4751545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4752972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4754401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4755874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4757612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4759066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4760501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4761918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4763348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4764770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4766195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4767599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4769064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4770552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4771959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4773383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4774852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4776337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4777745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4779175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4780588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4782022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 34%] 2024-08-06T21:15:47.4783428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4784865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4786287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4787766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4789216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4790627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4792064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4793534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4795023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4796470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4797899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4799360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4800778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4802177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4803602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4805614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4807176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4808650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4810066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4811501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4812955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4814436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4815878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4817664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4819097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4820520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4821935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4823374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4824776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4826296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4827767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4829209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4830634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4832085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4833559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4834972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4836404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4837802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4839235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4840709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4842128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4843531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4844994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4846478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4847947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4849353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4850767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4852242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4853690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4855110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4856558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4858001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4859406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4860835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4862250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4863721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4865171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4866929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4868416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4869848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4871298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4872745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4874170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4875589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4877030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4878429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4879857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4881271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4882725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4884171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4885596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4887031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4888446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4889887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4891361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4892805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4894218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4895638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4897071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4898551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4899953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4901412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4902864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4904750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4906171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4907600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4909124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4910611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4912027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4913852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4915312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4916726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4918157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4919556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4920985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4922490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4923984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4925394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4926834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4928321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4929795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4931211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4932638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4934078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4935490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4936917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4938375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4939810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4941248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4942724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4944138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4945573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4947023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4948502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4949913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4951332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4952747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4954155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4955589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4957010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4958444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4959920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4961431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4962856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4964281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4965751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4967283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4968705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4970126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4971537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4972941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4974373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4975771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4977214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4978673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4980165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4981566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4982992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4984403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4985879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4987343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4988768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4990187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4991601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4993021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4994431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4995874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4997356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.4998882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5000309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5001745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5003164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5005206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5006716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5008161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5009582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5010997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5012394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5013812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5015245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5016696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5018197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5019612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5021082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5022480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5024035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5025513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5026941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5028356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5029773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5031185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5032603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5034019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5035460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5036950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5038390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5039809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5041206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5042658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5044113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5045518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5046910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5048376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5049787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5051198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5052592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5053994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5055460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5056902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5058341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5059753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5061234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5062679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5064106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5065533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5066970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5068472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5069896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5071311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5072732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5074201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5075702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5077125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5078562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5080021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5081467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5082892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5084308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5085725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5087127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5088576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5089990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5091401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5092843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5094305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5095745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5097147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5098681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5100156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5101595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5102997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5104919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5106366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5107860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5109288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5110715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5112221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5113702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5115116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5116517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5118008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5119590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5121011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5122416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5123842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5125257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5127282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5130040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5132772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5135516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5138242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5140883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5143543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5146228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5149013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5151738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5154414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5157104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5159767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5162434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5165107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5167782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5170551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5173261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5175911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5178565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5181204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5183923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5186641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5189324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5191985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5194704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5197336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5200075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5202741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5206130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5208911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5211624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5214294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5216938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5219695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5222462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5225136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5227809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5230465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5233104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5235755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5238419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5241117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5243828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5246491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5249116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5251769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5254925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5257638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5260413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5263090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5265760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5268404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5271039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5273700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5276366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5279061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5281742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5284394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5287053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5289750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5292458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5295101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5297750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5300455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5303078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5306319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5308984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5311634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5314374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5317087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5319762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5322413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5325113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5327858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5330518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5333165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5335853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5338536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5341211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5343873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5346520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5349211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5351937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5354625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5357276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5359994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5362704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5365320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5367940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5370572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5373231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5375883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5378528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5381175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5383875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5386571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5389216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5391859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5394528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5397274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5400019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5402676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5405887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5408571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5411218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5414260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5417018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5419816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5422541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5425202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5427849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5430504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5433219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5435970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5438628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5441262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5443909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5446556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5449285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5451949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5454628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5457300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5459952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5462603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5465225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5467916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5470663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5473293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5475935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5478589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5481254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5483900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5486530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5489163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5491846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5494542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5497183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5499858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5502552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5505693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5508340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5510976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5513615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5516236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5518848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5521471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5524148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5526951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5529661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5532330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5534991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5537695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5540393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5543042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5545722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5548379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5551035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5553686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5556317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5558980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5561641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5564358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5567022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5569671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5572286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5574973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5577683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5580331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5582993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5585669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5588322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5590972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5593627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5596339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5599111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5601799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5604944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5607644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5610413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5613163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5615814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5618471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5621128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5623770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5626407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5629046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5631790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5634506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5637128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5639815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5642486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5645192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5647900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5650561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5653238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5655882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5658554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5661219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5663904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5666617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5669318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5671968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5674668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5677425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5681007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5683755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5686410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5689042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5691673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5694322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5697005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5699709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5702374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5705551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5708319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5710969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 35%] 2024-08-06T21:15:47.5713599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5716374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5719089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5721714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5724343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5726985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5729674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5732321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5734983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5737624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5740324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5743022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5745673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5748359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5751052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5753751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5756379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5759010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5761657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5764324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5766945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5769586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5772245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5774926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5777600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5780245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5782909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5785631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5788297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5791052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5793728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5796394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5799082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5801748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5804891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5807595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5810353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5813087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5815750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5818942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5821567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5824310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5827057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5829724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5832371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5835060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5837702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5840352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5842990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5845701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5848443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5851101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5853739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5856392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5859109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5861815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5864458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5867136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5870239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5872893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5875672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5878324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5881063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5883755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5886404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5889033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5891691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5894405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5897096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5899794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5902459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5905522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5908213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5910865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5913530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5916581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5919351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5922064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5924746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5927394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5930101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5932818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5935504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5938139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5940778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5943440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5946083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5948723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5951344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5954002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5956780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5959416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5962088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5964773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5967751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5970432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5973039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5975681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5978343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5981012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5983651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5986301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5989011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5991715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5994356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5997040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.5999777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6002475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6005682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6008350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6011015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6013658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6016287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6018947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6021591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6024361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6027094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6029743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6032406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6035055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6037760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6040470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6043126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6045772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6048455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6051161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6053872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6056538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6059238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6061936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6064606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6067258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6069881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6072582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6075293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6077937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6080566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6083202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6085869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6088484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6091111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6093762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6096457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6099183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6101806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6104891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6107664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6110368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6113034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6115679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6118337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6120968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6123614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6126263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6128924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6131643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6134345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6136966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6139696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6142374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6145044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6147698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6150372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6152993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6155625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6158289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6160983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6163631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6166343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6169108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6171757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6174417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6177043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6179733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6182446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6185074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6187731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6190375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6193035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6195735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6198377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6201132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6203810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6207005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6209644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6212279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6215029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6217733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6220342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6222979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6225642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6228281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6230906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6233541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6236260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6238959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6241586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6244258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6246955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6249650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6252309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6255355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6258046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6260683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6263350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6265978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6268585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6271219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6273894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6276580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6279226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6281869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6284518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6287214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6289856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6292486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6295106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6297762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6300460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6303102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6306201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6308948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6311686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6314339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6316959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6319880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6322623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6325357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6327988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6330863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6333526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6336129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6338769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6341400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6344099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6346788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6349466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6352102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6354745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6357436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6360130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6362786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6365481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6368147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6370816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6373505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6376218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6378940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6381666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6384329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6387054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6389740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6392454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6395195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6397908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6400603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6403258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6406479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6409183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6411861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6414643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6417789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6420482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6423159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6425825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6429074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6431966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6434727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6437403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6440094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6442798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6445481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6448159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6450898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6453645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6456329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6459002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6461695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6464426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6467142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6469792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6472477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6475161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6477825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6480512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6481938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6483387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6484863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6486352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6487769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6489208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6490654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6492129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6493552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6495009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6496435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6497845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6499329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6500765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6502190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6503685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6505622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6507062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6508478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6509953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6511444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6512866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6514296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6515726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6517140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6518575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6519978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6521402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6522876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6524430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6525865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6527303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6528772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6530278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6531691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6533129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6534555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6536011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6537444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6538864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6540300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6541760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6543234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6544647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6546106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6547575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6549044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6550457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6551894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6553314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6554744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6556180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6557607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6559049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6560761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6562343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6563776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6565226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6566700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6568185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6569615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6571070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6572489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6573928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6575347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6576963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6578925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6580406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6582581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6584738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6586899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6589544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6592432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6594498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6596087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6598651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6601579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6603767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6606421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6609137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6611885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6614904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6617646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6620288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6622831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6625413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6628097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6630376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6632483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6634540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6635983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6638516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6639977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6642308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6644223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6646691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6648549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6650541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6652219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6654263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6655879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6657670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6659610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6661047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6662536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6663966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6665464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6666933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6668374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6670255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6672006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6673524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6674991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6676456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6677867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6679298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6680713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6682151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6683557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6685020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6686502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6687941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6689345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6690754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6692233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6693719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6695140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6696579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6698020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6699485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6700915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6702324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6703807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6705759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6707215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6708632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6710061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6711567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6713036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6714461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6715882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6726553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6728134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6729581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6730997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6732435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6733975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6735494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6736932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6738365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6739812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6741286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6742710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6744156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6745564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6746994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6748432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6749859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6751287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6752739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6754215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6755629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6757063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6758506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6759979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6761395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6762820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6764221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6765651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 36%] 2024-08-06T21:15:47.6767087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6768493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6769914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6771370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6772854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6774258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6775679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6777179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6778666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6780069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6781497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6782913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6784391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6785799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6787245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6788653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6790103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6791561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6792963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6794387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6795805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6797275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6798770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6800583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6802035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6803450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6805356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6806822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6808260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6809756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6811652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6813359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6814818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6816223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6817743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6819242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6820681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6822086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6824061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6825504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6826939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6828331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6829794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6831275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6832694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6834107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6835508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6836977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6838476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6839891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6841295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6842726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6844140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6845849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6847618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6850014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6851544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6852953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6854407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6856258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6858001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6859500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6860952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6862367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6863799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6865194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6866618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6868033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6869467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6871322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6872802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6874232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6875717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6877178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6878642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6880071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6881505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6882926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6884329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6885770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6887187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6888608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6890055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6891556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6892982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6894388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6895865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6897399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6898874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6900279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6901721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6903134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6905042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6906476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6907901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6909404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6910906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6912327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6913732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6915243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6916727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6918578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6920022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6921463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6922905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6924331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6925745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6927185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6928687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6930169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6931576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6933026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6934480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6935922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6937336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6938750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6940181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6941578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6943020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6944442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6945876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6947415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6948893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6950302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6951730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6953147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6954584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6956059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6957476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6958895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6960295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6961720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6963161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6964575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6966014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6967481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6968889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6970291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6971780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6974321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6976809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6979457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6981441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6982921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6984337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6985811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6988422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6991400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6994324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.6997277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7000185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7003098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7006452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7009427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7012330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7015276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7018146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7020353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7022417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7024414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7026582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7028770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7030965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7032664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7034669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7036738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7038764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7040741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7042738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7044709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7046636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7048302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7050121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7052061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7053612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7055114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7056533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7057953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7059405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7060885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7062309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7063732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7065163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7066568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7067999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7069390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7070806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7072258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7073757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7075156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7076574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7078028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7079506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7080906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7082309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7083751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7085161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7086573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7087976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7089407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7091667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7093190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7094614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7096038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7097513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7099037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7100444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7101865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7103268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7105211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7106662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7108077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7109507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7111003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7112492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7113910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7115338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7116742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7118214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7119692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7121129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7122537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7123967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7125410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7126835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7128257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7129704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7131192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7132614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7134045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7135456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7136919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7138381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7139793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7141193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7142604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7144045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7145446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7146864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7148316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7149795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7151195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7152617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7154036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7155509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7156952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7158368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7159855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7161290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7162686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7164105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7165533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7166987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7168469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7169875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7171294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7172696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7174211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7175655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7177073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7178484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7179891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7181289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7182690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7184134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7185530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7186980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7188436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7189868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7191260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7192718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7194188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7195626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7197031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7198452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7199913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7201352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7202748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7204677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7206214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7207695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7209105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7210498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7211977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7213457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7214895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7216294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7217717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7219125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7220537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7221941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7223350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7224846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7226295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7227713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7229126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7230601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7232043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7233465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7234902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7236336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7237734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7239565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7241003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7242945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7245314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7246855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7248377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7249797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7251217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7252663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7254130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7255772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7257422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7258832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7260252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7261663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7263072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7264529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7265982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7267430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7268827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7270244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7271695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7273174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7274573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7275992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7277413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7278833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7280263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7281691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7283125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7284577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7285984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7287401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7288818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7290264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7291712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7293098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7294528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7295940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7297370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7298768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7300242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7301660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7303114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7305103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7306551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7308013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7309519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7311029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7312433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7313860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7315249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7316658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7318086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7319517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7320911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7322380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7323851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7325311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7326720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7328181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7329667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7331094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7332524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7333938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7335381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7336814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7338270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7339687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7341166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7342647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7344085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7345503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7346962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7348472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7349873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7351300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7352722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7354162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7355570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7357003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7358436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7359913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7361372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7362814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7364237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7365664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7367131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7368607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7370051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7371489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7372922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7374340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7375785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7377213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7378697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7380166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7381607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7383029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7384453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7385911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7387386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7388832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7390246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7391675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7393096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7394541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7395947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7397419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7398884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7400365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7401761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7403185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7405167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7406683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7408120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7409541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7410984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7412411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7413846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7415262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7416775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7418349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7420190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7421623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7423056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7424555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7426032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7427441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7428881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7430396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7431798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7433223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7434648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7436143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7437610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7439065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7440497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7441961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7443431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7444928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7446365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7447812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7449275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7450700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7452150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7453611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7455147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7456672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7458134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7459589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 37%] 2024-08-06T21:15:47.7461018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7462485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7463972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7465402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7466832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7468242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7469692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7471142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7472554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7473989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7475457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7477006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7478421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7479878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7481351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7482841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7484255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7485695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7487113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7488551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7489977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7491389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7492828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7494296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7495773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7497189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7498633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7500239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7501716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7503126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7505195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7506656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7508080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7509489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7510924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7512365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7513929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7515443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7516869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7518317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7519812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7521335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7522748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7524179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7525576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7526993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7528411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7529856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7531272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7532731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7534217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7535641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7537062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7538512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7540005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7541450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7542879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7544296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7545737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7547172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7548667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7550098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7551571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7553070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7554483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7555924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7557388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7558871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7560284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7561713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7563132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7564573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7565981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7567415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7568837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7570337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7571800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7573216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7574653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7576073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7577531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7578984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7580445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7581876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7583413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7584828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7586270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7587691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7589156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7590643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7592060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7593494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7594894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7596351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7597825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7599300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7600739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7602159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7603575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7605468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7606901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7608402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7609879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7611340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7612747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7614154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7615663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7617149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7618576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7619993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7621456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7622884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7624311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7625722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7627192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7628662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7630075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7631504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7632920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7634398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7635849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7637276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7638698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7640136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7641559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7642983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7644393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7645830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7647277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7648751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7650174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7651642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7653093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7654557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7655997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7657425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7658848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7660266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7661727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7663148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7664560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7666000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7667477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7668917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7670333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7671815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7673276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7674714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7676122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7677547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7678958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7680396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7681812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7683235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7684698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7686192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7687592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7689012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7690466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7691967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7693372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7694781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7696205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7697612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7699021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7700472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7701922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7703378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7705608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7707056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7708489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7709913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7711455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7712951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7714371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7715816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7717217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7718644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7720105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7721550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7723039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7724551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7725992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7727445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7728862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7730389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7731891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7733348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7734758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7736170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7737607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7739038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7740463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7741921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7743446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7744865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7746307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7747723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7749286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7750755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7752182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7753607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7755088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7756606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7758113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7759573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7761090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7762617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7764038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7765471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7766894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7768420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7769872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7771294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7772716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7774154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7775554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7776964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7778399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7779941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7781406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7782881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7784326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7785777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7787323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7788829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7790284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7791776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7793243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7794742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7796304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7797829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7799339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7801016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7802572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7804567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7806050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7807623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7809121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7810557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7811960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7813377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7814809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7816242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7817636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7819042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7820586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7822092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7823515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7824934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7826418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7827896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7829313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7830717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7832148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7833564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7835002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7836410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7837807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7839270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7840711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7842122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7843524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7845023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7846462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7847941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7849351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7850776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7852167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7853577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7855022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7856461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7857998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7859517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7860965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7862391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7863814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7865289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7866777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7868213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7869647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7871069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7872926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7874373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7875812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7877277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7878742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7880184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7881584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7883139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7884604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7886168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7887585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7889064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7890500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7891987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7893455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7894926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7896610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7898224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7899751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7901202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7902725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7905051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7906626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7908064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7909499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7910911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7912565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7913997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7915430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7916931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7918450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7920183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7921815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7923259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7924746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7926227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7927635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7929092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7930488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7931903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7933313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7934742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7936189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7937660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7939094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7940508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7941928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7943372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7944843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7946250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7947675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7949098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7950529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7951952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7953372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7954772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7956246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7957712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7959142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7960541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7962048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7963543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7964947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7966368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7967791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7969245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7970646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7972082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7973504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7974990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7976473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7977924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7979375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7980906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7982402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7983857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7985352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7986822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7988359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7989875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7991362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7992832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7994488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7996015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7997447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.7998862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8000346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8001807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8003285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8005313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8006761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8008186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8009619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8011063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8012466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8014060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8015563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8016997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8018397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8019814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8021289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8022771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8024187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8025590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8027016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8028432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8029859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8031263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8032733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8034227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8035640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8037037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8038473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8039955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8041406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8042829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8044246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8045678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8047078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8048505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8049917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8051387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8052821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8054341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8055746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8057174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8059429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8061005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8062410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8063828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8065240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8066634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8068061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8069471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8070897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8072337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8073821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8075240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8076660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8078109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8079697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8081134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8082542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8083959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8085360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8086784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8088172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8089586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8091050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8092524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8093920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8095334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8096778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8098247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8099697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8101148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8102555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8104388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8105948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8107351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8108781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8110342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8111857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8113259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8114683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8116146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8117613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8119011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8120427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8121849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8123241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8124646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 38%] 2024-08-06T21:15:47.8126045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8127465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8128886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8130339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8131760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8133185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8134571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8136024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8137477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8138912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8140309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8141730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8143158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8144574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8145996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8147488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8148971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8150387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8151825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8153227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8154678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8156128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8157534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8158932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8160357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8161785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8163184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8164600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8166037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8167509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8168900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8170314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8171739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8173214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8174667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8176079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8177492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8178920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8180315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8181738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8183165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8184577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8186056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8187506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8188925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8190328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8191794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8193228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8194644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8196053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8197458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8198847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8200285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8201739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8203130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8205022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8206533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8207965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8209360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8210844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8212332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8213774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8215174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8216594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8218003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8219432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8220835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8222265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8223719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8225172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8226576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8227969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8229390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8230833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8232309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8233712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8235134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8236538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8237944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8239344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8240745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8242230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8243666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8245079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8246484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8247909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8249355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8250811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8252239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8253672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8255116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8256549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8258340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8259776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8261175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8262635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8264106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8265507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8266914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8268340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8269799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8271201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8272626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8274015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8275436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8276849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8278241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8279658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8281104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8282603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8284006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8285426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8286876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8288357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8289752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8291174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8292591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8294013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8295395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8296807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8298208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8299699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8301158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8302573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8304374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8305853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8307347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8308811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8310237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8311657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8313098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8314503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8315935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8317352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8318800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8320303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8321712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8323166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8324566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8326020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8327465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8328886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8330272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8331682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8333104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8334535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8335921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8337348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8338815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8340226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8341632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8343045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8344511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8345960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8347367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8348763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8350182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8351589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8353027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8354424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8355849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8357291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8358729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8360141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8361534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8363008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8364435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8365835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8367233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8368650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8370033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8371437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8372859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8374280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8375702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8377138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8378559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8379974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8381462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8382928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8384365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8385787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8387199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8388603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8390028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8391447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8392881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8394336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8395789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8397210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8398597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8400053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8401511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8403006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8404780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8406229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8407635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8409060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8410453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8411866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8413358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8414830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8416232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8417622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8419043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8420515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8422386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8423844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8425272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8426679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8428086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8429479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8430883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8432275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8433743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8435183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8436582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8438003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8439430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8440882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8442281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8443720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8445104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8446507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8447953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8449382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8450767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8452214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8453676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8455086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8456492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8457926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8459407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8460811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8462219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8463621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8465033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8466428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8467832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8469213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8470666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8472112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8473504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8474910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8476301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8477756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8479187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8480593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8481993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8483419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8484802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8486210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8487607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8489066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8490503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8491900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8493329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8494741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8496191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8497623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8499038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8500480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8501880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8503267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8505135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8506577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8507983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8509459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8510926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8512346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8513747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8515215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8516673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8518094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8519482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8520889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8522285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8523704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8525090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8526498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8527937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8529410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8530845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8532232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8533700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8535153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8536548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8537931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8539341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8540734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8542130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8543520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8544937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8546369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8547820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8549205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8550603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8552022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8553446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8554938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8556339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8557772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8559158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8560563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8561964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8563383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8564823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8566283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8567676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8569071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8570465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8571892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8573354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8574772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8576173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8577558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8578970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8580370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8581767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8583154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8584640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8586108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8587509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8588926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8590387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8591874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8593281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8594724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8596139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8597577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8598982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8600448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8601858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8603328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8605185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8606635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8608051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8609545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8611032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8612442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8613875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8615307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8616724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8618131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8619568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8620985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8622461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8623933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8625379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8626803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8628254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8629728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8631148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8632592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8634028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8635460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8636868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8638304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8639697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8641153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8642621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8644069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8645473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8646894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8648345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8649810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8651228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8652637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8654073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8655491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8656914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8658317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8659791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8661261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8662678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8664098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8665533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8666989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8668442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8669866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8671270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8672692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8674094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8675512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8676915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8678390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8679866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8681284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8682689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8684138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8685576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8687022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8688458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8689882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8691301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8692708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8694168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8695595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8697060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8698518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8699998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8701428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8702848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8704773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8706304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8707720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8709123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8710539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8711949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8713385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8714800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8716229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8717693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8719187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8720591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8722011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8723469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8724972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8726374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8727786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8729220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8730643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8732068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8733477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8734935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8736400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8737878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8739290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8740712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8742166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8743646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8745071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8746503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8747964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8749374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8750797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8752208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8753641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8755103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8756592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8758005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8759437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8760875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8762341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8763756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8765216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8766617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8768024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8769456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8770878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8772299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8773738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8775229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 39%] 2024-08-06T21:15:47.8776637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8778042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8779434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8780899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8782360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8783773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8785189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8786610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8788020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8789417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8790831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8792285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8793768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8795192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8796614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8798033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8799541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8801000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8802430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8803848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8805701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8807109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8808523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8809948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8811433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8812912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8814312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8815759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8817171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8818662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8820127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8821561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8822982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8824401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8825823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8827253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8828670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8830111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8831577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8832992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8834640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8836174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8837662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8839123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8840558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8841959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8843378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8844776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8846218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8847606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8849038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8850508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8851915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8853323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8854719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8856201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8857656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8859067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8860469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8861895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8863306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8864719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8866130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8867546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8869018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8870465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8871888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8873650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8875175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8876624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8878046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8879447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8880870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8882259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8883669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8885080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8886511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8887948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8889418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8890844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8892256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8893714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8895172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8896605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8898017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8899430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8900870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8902300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8903718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8905619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8907136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8908623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8910056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8911453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8912875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8914333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8915817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8917206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8918618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8920019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8921446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8923204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8924664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8926145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8927636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8929030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8930429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8931851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8933296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8934755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8936158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8937590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8939006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8940420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8941823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8943248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8944699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8946163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8947570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8948966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8950382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8951809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8953256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8954655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8956094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8957487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8958894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8960292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8961715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8963104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8964566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8966035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8967449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8968861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8970302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8971777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8973197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8974620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8976036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8977469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8978881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8980297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8981703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8983205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8984668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8986085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8987478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8988927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8990397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8991789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8993202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8994606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8996045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8997440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.8998860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9000309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9001790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9003235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9005031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9006495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9008005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9009496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9010906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9012342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9013761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9015181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9016608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9018036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9019447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9021586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9023078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9024508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9025922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9027340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9028802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9030276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9031712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9033121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9034544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9035957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9037410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9038811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9040272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9041731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9043169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9044569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9045990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9047513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9049002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9050422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9051835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9053254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9054660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9056072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9057489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9058954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9060414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9061831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9063229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9064650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9066113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9067579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9068994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9070402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9071827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9073213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9074624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9076029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9077479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9078912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9080387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9081797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9083224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9084657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9086122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9087542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9088951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9090355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9091741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9093159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9094564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9095967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9097421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9098892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9100338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9101744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9103194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9105107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9106560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9107987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9109409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9110828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9112266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9113674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9115097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9116585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9118113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9119520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9120942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9122343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9123823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9125280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9126681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9128137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9129556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9130973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9132378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9133807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9135259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9136725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9138153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9139582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9141001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9142459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9143908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9145347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9146774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9148199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9149627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9151047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9152485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9153936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9155410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9156818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9158269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9159658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9161119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9162569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9164005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9165408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9166809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9168257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9169675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9171086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9172525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9174013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9175421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9176829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9178248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9179718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9181177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9182587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9183986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9185388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9186816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9188231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9189648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9191046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9192510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9193942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9195353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9196754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9198244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9199716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9201136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9202535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9204275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9205773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9207165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9208625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9210043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9211537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9213008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9214440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9215856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9217330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9218814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9220245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9221670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9223082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9224486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9225886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9227314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9228721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9230170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9231628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9233062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9234458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9235875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9237318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9238819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9240215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9241638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9243043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9244450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9245856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9247248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9248749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9250216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9251626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9253024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9254448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9255895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9257352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9258936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9260585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9262000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9263400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9264795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9266197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9267681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9269144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9270558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9271956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9273382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9274810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9276335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9277736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9279185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9280575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9281983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9283390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9284799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9286213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9287647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9289139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9290555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9291971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9293411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9294871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9296270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9297680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9299089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9300553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9301963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9303346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9305225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9306760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9308257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9309669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9311080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9312538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9314041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9315425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9316835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9318237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9319682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9321070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9322488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9323896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9325373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9326842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9328242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9330130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9331567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9333036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9334473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9335888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9337296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9338702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9340116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9341537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9342941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9344369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9345828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9347232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9348704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9350113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9351577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9353024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9354448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9355839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9357246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9358644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9360094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9361484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9362911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9364363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9365758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9367148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9368525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9369997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9371444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9372840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9374227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9375642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9377036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9378438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9379846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9381247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9382710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9384154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9385567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9386975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9388509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9389989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9391404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9392810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9394238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9395633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9397050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9398452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9399935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9401366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9402808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9404666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9406108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9407616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9409075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9410516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9411921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9413333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9414724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9416149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9417565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9418973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9420465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9421945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9423662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9425239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9426668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9428143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9429624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9431041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 40%] 2024-08-06T21:15:47.9432460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9433855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9435276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9436669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9438081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9439519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9440996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9442399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9443787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9445204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9446650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9448099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9449486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9450936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9452347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9453750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9455144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9456565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9457980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9459430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9460915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9462320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9463743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9465174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9466635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9468031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9469451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9470853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9472261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9473660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9475081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9476466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9477912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9479361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9480795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9482201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9483635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9485092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9486493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9487894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9489277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9490716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9492119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9493528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9494917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9496369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9497824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9499217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9500694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9502091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9503541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9505460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9506893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9508301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9509718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9511127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9512539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9513936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9515455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9516908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9518293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9519718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9521151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9522638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9524105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9525532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9526955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9528370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9529773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9531226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9532644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9534117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9535568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9536988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9538394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9539780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9541255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9542711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9544141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9545534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9546945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9548344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9549771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9551180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9552589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9554032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9555511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9556901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9558299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9559769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9561250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9562670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9564072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9565498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9566910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9568319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9569719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9571157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9572604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9580632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9582159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9583596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9585184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9586668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9588083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9589509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9590920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9592313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9593725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9595132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9596559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9598014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9599476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9600956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9602385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9603773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9605785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9607301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9608756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9610151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9611548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9612965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9614370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9615777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9617224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9618724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9620131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9621536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9622927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9624389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9625835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9627247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9628667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9630094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9631504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9632898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9634313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9635721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9637198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9638670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9640093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9641505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9642979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9644428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9645847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9647251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9648751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9650150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9651541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9652957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9654364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9655813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9657271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9658721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9660128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9661583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9663023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9664443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9665847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9667253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9668668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9670071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9671498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9672891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9674343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9675801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9677235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9678648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9680059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9681509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9682973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9684354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9685756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9687155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9688583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9689990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9691378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9692835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9694284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9695684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9697075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9698514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9700003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9701465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9702858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9704777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9706238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9707651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9709075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9710479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9712002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9713460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9714869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9716263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9717677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9719136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9720607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9722003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9723426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9724809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9726211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9727606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9729030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9730434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9731866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9733336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9734742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9736145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9737582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9739070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9740479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9741883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9743274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9744695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9746100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9747495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9748927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9750365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9751842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9753220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9754620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9756014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9757470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9758919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9760326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9761724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9763141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9764520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9765910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9767319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9768775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9770230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9771614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9773040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9774443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9775889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9777322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9778762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9780168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9781567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9782953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9784346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9785756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9787173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9788626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9790023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9791439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9792815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9794252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9795682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9797095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9798488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9799920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9801320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9802740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9804848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9806387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9807980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9809481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9810893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9812287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9813764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9815235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9816637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9818025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9819440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9820839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9822236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9823625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9825023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9826483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9827915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9829330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9830729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9832150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9833571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9835022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9836471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9837941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9839383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9840831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9842283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9843734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9845221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9846712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9848178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9849645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9851097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9852574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9854094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9855530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9856973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9858396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9859854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9861297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9862739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9864210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9865718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9867177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9868617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9870063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9871543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9873055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9874714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9876190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9877630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9879116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9880553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9882004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9883495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9885006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9886436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9887869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9889344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9890821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9892301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9893763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9895228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9896672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9898114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9899565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9901023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9902503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9904362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9905956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9907415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9908856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9910387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9911897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9913335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9914785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9916209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9917656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9919094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9920563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9922145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9924458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9925921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9927373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9928837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9930505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9931960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9933399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9934831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9936251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9937697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9939137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9940583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9942067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9943580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9945027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9946468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9947942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9949457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9950926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9952376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9953811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9955255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9956722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9958154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9959606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9961103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9962611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9964040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9965483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9966961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9968469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9969892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9971363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9972801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9974241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9975684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9977119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9978575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9980061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9981579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9983015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9984475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9985966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9987463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9988901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9990362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9991821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9993310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9994746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9996183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9997639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:47.9999100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0000589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0002030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0003486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0005443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0007009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0008450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0009916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0011359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0012803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0014236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0015694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0017116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0018596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0020118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0021583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0023034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0024506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0026006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0027449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0028894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0030327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0031794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0033230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0034661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0036075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0037562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0039045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0040466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0041932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0043407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0044900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0046318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0047760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0049201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0050662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0052112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0053567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0055012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0056516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0057997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0059451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0060895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0062410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0063903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0065341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0066792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0068232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0069678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0071104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0072570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0074012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0075492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0076972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0078427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0079868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0081383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0082862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0084298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0085779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0087205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0088647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0090082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0091557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0093025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0094524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 41%] 2024-08-06T21:15:48.0096013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0097473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0098900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0100437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0101933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0103387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0105274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0106733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0108184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0109625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0111066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0112511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0114048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0115558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0116991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0118410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0119920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0121423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0122886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0124315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0125750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0127204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0128630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0130070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0131500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0133022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0135849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0138547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0141226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0143972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0146707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0149401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0152117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0154859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0157529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0160227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0162910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0165611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0168344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0171099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0173794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0176483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0179235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0181980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0184678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0187433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0190140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0192853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0195564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0198280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0200979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0203724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0206986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0209680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0212363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0215128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0217886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0220591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0223289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0225992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0228669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0231379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0234100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0236784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0239534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0242305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0244997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0247676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0250422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0253181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0255880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0258580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0261617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0264371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0267068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0269760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0272457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0275201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0277917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0280600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0283306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0286044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0288799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0291471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0294198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0296879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0299562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0302258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0305411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0308184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0310984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0313727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0316445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0319155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0321952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0324706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0327408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0330101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0332804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0335525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0338218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0340929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0343610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0346299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0349038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0351735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0354416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0357144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0359877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0362598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0365268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0367954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0370650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0373359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0376059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0378743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0381485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0384247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0386946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0389651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0392397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0395220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0397896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0400586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0403274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0406459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0409173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0411843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0414534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0417314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0420072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0422756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0425447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0428213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0430979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0433673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0436397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0439109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0441805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0444500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0447210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0449912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0452682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0455450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0458157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0460872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0463642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0466361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0469060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0471791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0474476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0477153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0479844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0482582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0485278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0487989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0490755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0493498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0496195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0498933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0501679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0504798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0507520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0510225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0512906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0515606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0518296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0520956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0523733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0526506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0529223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0531902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0534639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0537365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0540029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0542688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0545373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0548067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0550742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0553410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0556088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0558813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0561537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0564228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0566912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0569673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0572399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0575081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0577773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0580482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0583230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0585914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0588603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0591320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0594101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0596842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0599532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0602265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0605413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0608179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0610885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0613587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0616282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0618966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0621679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0624372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0627073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0629831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0632645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0635355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0638050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0640789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0643531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0646245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0648944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0651633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0654335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0657055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0659772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0662464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0665201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0667948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0670629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0673302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0676027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0678781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0681471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0684167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0686852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0689553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0692225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0694965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0697659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0700400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0703168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0706319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0708996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0711795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0714561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0717242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0719930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0722629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0725306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0727973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0730692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0733394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0736122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0738851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0741529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0744200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0746921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0749648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0752343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0755037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0757732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0760377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0763065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0765779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0768472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0771220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0773967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0776682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0779380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0782109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0784876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0787586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0790303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0793013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0795706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0798393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0801081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0803777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0806895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0809689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0812362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0815045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0817797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0820560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0823250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0825930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0828620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0831342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0834019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0836695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0839395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0842154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0844880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0847577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0850282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0853040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0855777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0858473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0861142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0863854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0866531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0869192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0871879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0874569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0877312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0880011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0882703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0885419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0888141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0890854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0893579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0896261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0898947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0901636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0904791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0907535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0910228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0913005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0915778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0918476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0921183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0923943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0926688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0929355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0932044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0934703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0937382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0940095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0942780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0945442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0948139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0950879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0953558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0956217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0958967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0961704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0964435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0967105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0969801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0972509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0975205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0977931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0980607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0983349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0986107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0988785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0991460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0994186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0996912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.0999604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1002286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1005371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1008072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1010740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1013422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1016106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1018857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1021595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1024277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1026977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1029648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1032381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1035146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1037842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1040526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1043200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1045879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1048561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1051243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1053960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1056683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1059391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1062058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1064695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1067406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1070174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1072851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1075520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1078201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1080890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1083551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1086244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1088975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1091731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1094470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1097146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1099822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1102586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1105733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1108446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1111150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1113868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1116538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1119222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1121898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1124663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1127420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1130084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1132761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1135453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1138188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1140936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1143608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1146321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1148997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1151636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1154332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1157054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1159827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1162570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1165298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1168015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1170727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1173476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1176245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1178971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1181685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1184381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1187102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1189808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1192504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1195332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1198087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1200802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1203515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1206597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1209379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1212155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1214861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1217554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1220240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1222976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1225687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1228387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1231156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1233987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1236684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1239396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1242120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1244890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1247644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1250355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1253073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1255785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1258509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1261220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1263925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1266680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1269421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1272091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1274789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1277525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1280293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1283031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1285730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1288429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1291123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1293852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1296583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 42%] 2024-08-06T21:15:48.1299311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1302049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1305179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1307931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1310637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1313336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1316121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1318876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1321574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1324241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1326944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1329630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1332334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1335042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1337771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1340507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1343207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1345899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1348580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1351326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1354127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1356816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1359512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1362225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1364948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1367667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1370384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1373166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1375922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1378632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1381347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1384053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1386800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1389547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1392213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1394955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1397673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1400398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1403098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1406202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1408993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1411762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1414455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1417166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1419897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1422662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1425402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1428165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1430906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1433618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1436321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1439041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1441770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1444503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1447304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1450027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1452742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1455443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1458176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1460909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1463625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1466368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1469063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1471767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1474503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1477187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1479920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1482676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1485400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1488098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1490834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1493658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1496356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1499062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1500509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1501947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1503404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1505244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1506722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1508228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1509733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1511173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1512593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1514107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1516323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1517760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1519183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1520627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1522075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1523518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1524941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1526398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1527887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1529381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1530811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1532274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1533783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1535256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1536713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1538155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1539617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1541047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1542524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1543957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1545413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1546880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1548371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1549806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1551261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1552749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1554222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1555672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1557113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1558551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1559978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1561427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1562892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1564330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1565790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1567294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1568734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1570171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1571653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1573164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1574606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1576031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1577477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1578909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1580355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1581795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1583236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1584709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1586209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1587628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1589086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1590558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1592065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1593531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1594962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1596412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1597850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1599289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1600717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1602194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1603676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1605589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1607041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1608495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1610013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1611524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1612978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1614428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1615860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1617276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1618711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1620145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1621598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1623101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1624608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1626039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1627494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1628958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1630502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1631938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1633423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1634843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1636283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1637718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1639163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1640607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1642074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1643611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1645053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1646498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1647967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1649453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1650881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1652318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1653749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1655196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1656629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1658071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1659493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1660972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1662472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1663893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1665329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1666804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1668300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1669716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1671159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1672600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1674059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1675481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1676929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1678367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1679850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1681343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1682785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1684226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1685697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1687170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1688590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1690039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1691471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1692959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1694387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1695835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1697267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1698742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1700211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1701654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1703135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1705043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1706581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1708023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1709484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1710907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1712355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1713815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1715280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1716711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1718222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1719717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1721166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1722581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1724059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1725558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1726992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1728432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1729855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1731302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1732737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1734194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1735612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1737104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1738593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1740034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1741455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1742952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1744456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1745883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1747327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1748762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1750220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1751643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1753092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1754524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1756016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1757472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1758908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1760339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1761836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1763308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1764732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1766180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1767608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1769045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1770464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1771918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1773375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1774856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1776327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1777785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1779221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1780703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1782187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1783666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1785108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1786531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1787981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1789407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1790861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1792278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1793817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1795299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1796755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1798174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1799658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1801134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1802589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1804350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1805830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1807258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1808688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1810124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1811542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1813056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1814582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1816024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1817445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1818945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1820441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1821881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1823308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1824764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1826188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1827616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1829035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1830467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1831959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1833420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1834872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1836298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1837788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1839250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1840681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1842116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1843574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1845001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1846445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1847892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1849333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1850811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1852291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1853755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1855198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1856697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1858174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1859624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1861057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1862489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1863932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1865384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1866822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1868257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1869721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1871202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1872656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1874099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1875589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1877071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1878529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1879948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1881390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1882828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1884313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1885746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1887189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1888667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1890176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1891601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1893069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1894587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1896064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1897497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1898914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1900365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1901801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1903237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1905109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1906587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1908098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1909601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1911024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1912468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1913961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1915453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1916891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1918325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1919779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1921195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1922630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1924067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1925534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1926990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1928490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1929909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1931349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1932759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1934274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1935750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1937177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1938602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1940012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1941444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1942871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1944306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1945764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1947267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1948702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1950137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1951559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1953047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1954553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1955980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1957439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1958876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1960333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1961760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1963206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1964696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1966195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1967610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1969039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1970469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1971960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1973421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1974884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1976318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1977750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1979185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1980608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1982052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1983523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1985025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1986441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1987889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1989322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1990792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1992253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1993738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1995199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1996635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1998053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.1999473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2000910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2002368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2003843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2005633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2007079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2008487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2009991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2011479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2012924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2014336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2015782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2017204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2018657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2020070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2021545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2023057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2024492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2025926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2027344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2028831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2030315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2031758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2033186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2034632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2036058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2037484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2038898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2040360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2041851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2043260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2044703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2046127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2047607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2049060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2050491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2051920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2053365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2054801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 43%] 2024-08-06T21:15:48.2056238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2057666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2059163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2060632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2062048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2063487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2064945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2066423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2067907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2069348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2070772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2072193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2073606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2075077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2076508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2077971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2079438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2080881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2082306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2083713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2085208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2086680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2088129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2089542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2090984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2092413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2093908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2095351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2096830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2098302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2099752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2101167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2102591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2104465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2106035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2107461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2108882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2110322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2111750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2113179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2114595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2116112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2117599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2119023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2120435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2121880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2123358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2124821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2126274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2127709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2129163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2130578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2132026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2133453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2134943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2136416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2137854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2139272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2140712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2142177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2143636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2145092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2146525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2147952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2149362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2150807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2152237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2153705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2155173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2156624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2158046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2159476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2160936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2162432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2163863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2165296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2166732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2168157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2169609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2171026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2172504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2173970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2175436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2176842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2178269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2179733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2181227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2182636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2184074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2185521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2186947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2188374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2189788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2191236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2192710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2194230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2195672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2197119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2198592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2200073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2201493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2202942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2204765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2206238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2207677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2209098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2210540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2212014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2213527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2214951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2216419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2217888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2219382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2220807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2222255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2223668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2225100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2226545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2227972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2229396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2230845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2232381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2233815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2235245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2236720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2238209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2239636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2241064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2242476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2243908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2245324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2246744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2248165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2249621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2251106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2252515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2253939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2255393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2256894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2258295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2259721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2261148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2262596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2264009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2265447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2266893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2268367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2269848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2271272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2280902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2282494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2284032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2285532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2286959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2288368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2289777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2291178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2292603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2294098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2295578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2297038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2298457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2299869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2301280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2302723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2304683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2306328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2307748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2309169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2310592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2312030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2313439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2314974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2316471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2317909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2319312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2320729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2322200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2323677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2325100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2326510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2327932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2329350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2330763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2332168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2333624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2335104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2336513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2337918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2339332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2340786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2342234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2343629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2345065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2346491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2347892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2349310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2350726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2352188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2353635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2355067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2356467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2357874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2359306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2360769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2362177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2363595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2365020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2366419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2367834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2369240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2370683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2372125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2373544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2375007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2376419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2377870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2379340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2380766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2382188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2383601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2385043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2386480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2387887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2389301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2390754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2392224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2393663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2395096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2396553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2398017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2399415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2400829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2402242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2403653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2405609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2407046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2408465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2409985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2411471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2412872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2414288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2415778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2417246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2418651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2420070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2421484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2422891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2424295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2425707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2427118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2428549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2430009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2431420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2432837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2434281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2435749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2437156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2438572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2439961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2441364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2442770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2444186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2445598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2447034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2448499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2449914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2451319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2452768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2454233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2455654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2457062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2458464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2459866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2461269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2462664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2464072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2465523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2466993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2468388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2469791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2471192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2472648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2474098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2475504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2476916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2478332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2479726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2481137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2482551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2486868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2490146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2491570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2492986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2494493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2495909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2497317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2498733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2500145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2501534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2502928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2504821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2506262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2507771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2509372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2510859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2512268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2513671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2515065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2516472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2517882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2519305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2520699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2522109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2523532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2524919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2526370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2527885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2529383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2530771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2532171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2533571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2534976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2536362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2537761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2539177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2540581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2541977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2543366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2544764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2546291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2547775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2549185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2550599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2552016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2553410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2554803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2556225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2557638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2559063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2560471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2561882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2563302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2564780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2566273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2567675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2569110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2570502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2571891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2573291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2574710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2576099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2577499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2578929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2580342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2581732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2583213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2584733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2586160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2587578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2589018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2590451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2591879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2593346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2594765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2596191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2597623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2599065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2600479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2601968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2603477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2605348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2606803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2608235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2609682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2611089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2612509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2613926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2615349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2616759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2618180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2619612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2621194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2622727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2624140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2625574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2627011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2628436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2629855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2631286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2632715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2634131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2635548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2636976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2638405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2639887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2641394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2642808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2644237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2645650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2647061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2648496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2649927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2651334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2652743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2654158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2655585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2656981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2658491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2659991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2661421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2662832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2664246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2665669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2667095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2668535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2669942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2671354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2672767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2674176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2675571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2677060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2678581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2679981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2681388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2682814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2684226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2685624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2687033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2688476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2689905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2691314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2692738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2694199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2695716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2697217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2698650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2700073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2701515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2702923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2704584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2706025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2707448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2708857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2710266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2711700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2713122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2714645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2716179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2717606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2719032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2720451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2721860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 44%] 2024-08-06T21:15:48.2723286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2724718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2726131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2727554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2728983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2730422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2731830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2733326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2734838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2736270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2737707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2739136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2740550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2741983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2743393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2744810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2746224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2747675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2749094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2750500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2751999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2753504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2754915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2756325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2757779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2759196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2760606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2762012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2763435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2764852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2766269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2767713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2769131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2770639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2772132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2773547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2774961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2776385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2777800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2779210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2780626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2782064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2783471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2784901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2786328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2787785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2789279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2790779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2792223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2793701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2795137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2796563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2798026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2799465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2800905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2802321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2803763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2805697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2807156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2808731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2810279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2811722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2813131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2814565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2815991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2817450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2818873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2820302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2821725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2823172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2824596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2826026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2827540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2829104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2830510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2831919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2833367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2834798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2836235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2837681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2839129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2840558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2841988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2843414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2844840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2846335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2847860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2849267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2850696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2852128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2853534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2854956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2856381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2857844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2859250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2860671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2862097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2863530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2865025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2866534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2867979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2869426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2870850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2872280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2873704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2875146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2876572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2878001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2879435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2880864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2882279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2883757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2885274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2886696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2888135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2889556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2890986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2892399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2893850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2895294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2896716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2898156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2899567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2900990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2902497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2904291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2905737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2907171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2908604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2910045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2911452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2912879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2914297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2915708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2917135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2918550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2919979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2921517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2923062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2924476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2925897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2927347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2928770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2930177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2931616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2933044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2934461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2935879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2937332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2938774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2940226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2941741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2943212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2944666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2946080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2947527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2948933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2950359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2951759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2953174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2954595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2956022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2957458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2958982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2960495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2961961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2963378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2964794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2966223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2967667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2969093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2970513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2971941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2973371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2974804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2976216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2977706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2979236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2980699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2982119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2983540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2984977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2986373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2987807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2989224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2990654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2992062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2993524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2994939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2996405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2997933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.2999396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3000826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3002258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3003688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3005554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3007023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3008473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3009899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3011319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3012762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3014191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3015697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3017239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3018719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3020156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3021563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3022992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3024410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3025851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3027273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3028705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3030128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3031569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3032978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3034443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3035950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3037501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3038912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3040327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3041775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3043210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3044641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3046070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3047534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3048959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3050380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3051797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3053257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3054760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3056222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3057650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3059086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3060514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3061919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3063338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3064763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3066198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3067630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3069057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3070469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3071943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3073446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3074909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3076321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3077774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3079188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3080596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3082025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3083446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3084854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3086264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3087718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3089137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3090548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3092018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3093568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3094986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3096402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3097828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3099250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3100671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3102074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3103478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3105166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3106618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3108042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3109471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3111007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3112564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3113973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3115409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3116841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3118301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3119719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3121137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3122567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3123983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3125404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3126827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3128252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3129746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3131384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3132790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3134226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3135653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3137075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3138487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3139930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3141360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3142769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3144202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3145626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3147078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3148575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3150086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3151504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3152941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3154364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3155784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3157216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3158645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3160049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3161468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3162888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3164311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3165722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3167226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3168738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3170162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3171586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3173005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3174433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3175850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3177284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3178693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3180113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3181540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3182955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3184368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3185855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3187384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3188786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3190212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3191623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3193043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3194475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3195889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3197320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3198746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3200146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3201563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3202967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3204998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3206564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3207990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3209432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3210868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3212291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3213706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3215143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3216574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3218009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3219423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3220864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3222290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3223787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3225291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3226701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3228143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3229549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3230967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3232383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3233817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3235221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3236651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3238073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3239517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3240924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3242416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3243914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3245334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3246756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3248171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3249597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3251018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3252437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3253843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3255266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3256701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3258120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3259528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3261020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3262516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3263924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3265322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3266757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3268176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3269568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3270975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3272393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3273821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3275226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3276670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3278082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3279559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3281063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3282527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3283950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3285411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3286848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3288265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3289715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3291152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3292582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3294053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3295501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3296944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3298411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3299906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3301395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3302815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3304493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3305931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3307370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3308814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3310228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3311654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3313077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3314532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3315936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3317453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3318994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3320502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3321912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3323345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3324779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3326225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3327658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3329086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3330519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3331941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3333370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3334770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3336243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3337770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3339242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3340645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3342082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3343512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3344933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3346339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3347777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3349210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3350616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3352047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3353473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3354957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3356500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3357988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3359409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3360855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3362267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3363691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3365102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3366541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3367949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3369354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3370787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3372205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3373619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3375094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3376607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3378025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 45%] 2024-08-06T21:15:48.3379449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3380864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3382309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3383734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3385167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3386596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3388024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3389471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3390891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3392324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3393863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3395394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3396817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3398255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3399675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3401105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3402507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3404254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3405884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3407345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3408761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3410176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3411608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3413158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3414693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3416101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3417566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3419005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3420429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3421845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3423295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3424726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3426151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3427594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3429032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3430475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3432027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3433561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3434978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3436423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3437858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3439284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3440706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3442151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3443558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3444981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3446403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3447876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3449282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3450762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3452286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3453711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3455143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3456561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3458030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3459457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3460894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3462315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3463753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3465184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3466620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3468054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3469561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3471063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3472467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3473894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3475318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3476761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3478184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3479616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3481040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3482477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3483886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3485325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3486736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3488262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3489766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3491169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3492606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3494071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3495503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3496915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3498367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3499800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3501223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3502638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3504320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3505752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3507295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3508829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3510323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3511740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3513149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3514572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3515982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3517414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3518832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3520250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3521676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3523124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3524531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3526004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3527529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3529025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3530438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3531859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3533307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3534736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3536171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3537604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3539044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3540467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3541891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3543295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3544775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3546284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3547773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3549188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3550627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3552059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3553468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3554901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3556330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3557796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3559209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3560644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3562071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3563563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3565080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3566566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3568012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3569460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3570882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3572293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3573726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3575150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3576571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3577995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3579440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3580872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3582295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3583774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3585297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3586713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3588163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3589581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3591010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3592435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3593879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3595304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3596723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3598194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3599604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3601033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3602531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3604475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3605932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3607364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3608803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3610236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3611630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3613035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3614468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3615886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3617305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3618749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3620179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3621724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3623267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3624670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3626116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3627550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3628999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3630419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3631865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3633298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3634708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3636141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3637571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3639033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3640541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3642062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3643483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3644919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3646323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3647746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3649185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3650631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3652037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3653465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3654896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3656329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3657751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3659248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3660763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3662185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3663607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3665022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3666459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3667884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3669325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3670734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3672169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3673597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3675016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3676446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3677932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3679453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3680850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3682265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3683684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3685121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3686521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3687944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3689364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3690800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3692203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3693666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3695087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3696553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3698067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3699523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3700963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3702396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3703823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3705516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3706964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3708407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3709832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3711250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3712685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3714095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3715569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3717127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3718633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3720068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3721477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3722904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3724313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3725749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3727154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3728591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3730010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3731454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3732859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3734373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3735887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3737358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3738795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3740212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3741656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3743075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3744498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3745915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3747341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3748776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3750205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3751606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3753028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3754522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3756008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3757427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3758865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3760299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3761693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3763111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3764525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3765961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3767364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3768816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3770238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3771673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3773144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3774647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3776081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3777510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3778960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3780369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3781794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3783214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3784624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3786019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3787444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3788892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3790312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3791788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3793298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3794752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3796161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3797589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3799036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3800482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3801895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3803321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3805196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3806685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3808103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3809547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3811093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3812644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3814051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3815462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3816895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3818311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3819747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3821152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3822594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3824011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3825431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3826843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3828271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3829784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3831284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3832686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_1024_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3834136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3835576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3836990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3838423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3839883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3841348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3842769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3844220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3845662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3847115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3848576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3850140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3851610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3853060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3854479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3855912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3857340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3858777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3860220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3861634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3863080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3864518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3865955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3867418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3868960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3870444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3871888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3873319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3874775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3876210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3877636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3879098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3880531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3881986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3883416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3884861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3886363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3887897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3889326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3890764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3892195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3893689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3895110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3896548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3897983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3899442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3900880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3902308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3903750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3905476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3907017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3908492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3909959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3911400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3912831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3914243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3915691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3917137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3918571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3920013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3921443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3922882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3924328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3925861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3927331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3928779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3930213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3931649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3933061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3934507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3935925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3937356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3938793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3940267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3941686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3943154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3944694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3946180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3947620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3949054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3950514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3951947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3953387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3954820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3956260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3957690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3959135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3960553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3962022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3963563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3965032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3966472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3967906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3969372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3970785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3972222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3973656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3975110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3976532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3977975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3979427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3980929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3982439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3983912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3985368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3986811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3988253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3989697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3991146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3992583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3994047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3995470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3996921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3998352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.3999861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4001372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4002872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4004781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4006236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4007678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4009099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4010569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4011985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4013418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4014850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4016302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4017715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4019235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4020809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4022386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4023810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4025258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4026683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4028102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4029536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4030971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4032411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4033838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4035273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 46%] 2024-08-06T21:15:48.4036687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4038168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4039685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4041168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4042586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4044045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4045490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4046911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4048352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4049803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4051260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4052689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4054139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4055573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4057074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4058586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4060112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4061543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4062992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4064408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4065835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4067270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4068714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4070169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4071597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4073049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4074482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4075960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4077470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4078967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4080427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4081863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4083280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4084721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4086155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4087594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4089014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4090475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4091938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4093354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4094882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4096395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4097886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4099287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4100739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4102169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4103610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4105283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4106732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4108150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4109595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4111030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4112441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4113955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4115503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4116996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4118416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4119866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4121322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4122753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4124175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4125630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4127062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4128500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4129933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4131364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4132852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4134367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4135843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4137266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4138716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4140135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4141566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4142991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4144444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4145853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4147286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4148715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4150177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4151643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4153150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4154643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4156082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4157521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4158949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4160411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4161856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4163296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4164715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4166151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4167581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4169013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4170489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4172019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4173490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4174907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4176341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4177771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4179215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4180650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4182090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4183519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4184965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4186382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4187820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4189287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4190849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4192310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4193768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4195227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4196668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4198099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4199518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4201003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4202427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4203849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4205724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4207179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4208600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4210161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4211706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4213142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4214565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4215982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4217410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4218842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4220296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4221729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4223163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4224595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4226048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4227459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4228971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4230483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4231949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4233368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4234812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4236230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4237653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4239080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4240491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4241948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4243380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4244809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4246220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4247730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4249248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4250673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4252091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4253541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4254969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4256378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4257813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4259242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4260756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4262172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4263607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4265029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4266551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4268050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4269481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4270913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4272359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4273762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4275186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4276611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4278033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4279481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4280922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4282362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4283781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4285287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4286790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4288227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4289659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4291111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4292523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4294006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4295454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4296896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4298315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4299747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4301223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4302637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4304404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4306000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4307438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4308848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4310279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4311720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4313160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4314574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4316013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4317428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4318853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4320284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4321716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4323237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4324789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4326208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4327615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4329060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4330479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4331922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4333336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4334772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4336191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4337618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4339040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4340448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4341971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4343467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4344885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4346298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4347737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4349142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4350565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4351992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4353426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4354822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4356246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4357677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4359098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4360596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4362094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4363538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4364971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4366409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4367830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4369272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4370714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4372155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4373566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4375000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4376425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4377846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4379323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4380841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4382285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4383693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4385124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4386543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4387986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4389399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4390848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4392267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4393742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4395164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4396595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4398732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4400298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4401734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4403158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4405069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4406541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4407971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4409391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4410831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4412270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4413690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4415106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4416539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4418091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4419649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4421064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4422532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4423965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4425376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4426808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4428377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4429833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4431244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4432692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4434120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4435560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4437018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4438540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4440004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4441449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4442887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4444306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4445733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4447154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4448571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4449970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4451404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4452853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4454265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4455709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4457223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4458688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4460112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4461519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4462975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4464399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4465810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4467247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4468670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4470118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4471540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4472977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4474448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4475985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4477445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4478881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4480301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4481744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4483143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4484552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4485995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4487412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4488839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4490257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4491698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4493164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4494726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4496190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4497626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4499047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4500472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4501893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4503334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4504981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4506400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4507831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4509259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4510697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4512186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4513734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4515206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4516640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4518038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4519463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4520875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4522338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4523748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4525170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4526584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4528008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4529434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4530882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4532425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4533892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4535312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4536722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4538161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4539582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4541007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4542445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4543882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4545309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4546715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4548146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4549602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4551125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4552593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4554010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4555428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4556869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4558268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4559694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4561113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4562565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4563966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4565396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4566813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4568232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4569742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4571234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4572691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4574117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4575548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4576954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4578388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4579821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4581247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4582674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4584107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4585524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4586910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4588402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4589899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4591330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4592749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4594216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4595630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4597066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4598473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4599901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4601311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4602750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4604581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4606073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4607615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4609148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4610567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4611991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4613434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4614857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4616281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4617699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4619129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4620538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4621962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4623365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4624788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4626331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4627832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4629251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4630664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4632115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4633519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4634939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4636367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4637805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4639204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4640631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4642067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4643507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4644990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4646498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4647918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4649342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4650769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4652201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4653630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4655048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4656463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4657864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4659296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4660724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4662157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4663639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4665159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4666569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4667987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4669400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4670836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4672315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4673744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4675193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4676625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4678087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4679521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4680963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4682493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4684038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4685460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4686908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4688344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4689781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4691208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4692648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4694139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 47%] 2024-08-06T21:15:48.4695580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4697019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4698447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4699895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4701410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4702978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4704664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4706141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4707590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4709033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4710452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4711890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4713367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4714788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4716240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4717685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4719143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4720674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4722240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4723668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4725115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4726529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4727966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4729393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4730850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4732280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4733702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4735147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4736586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4738022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4739517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4741044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4742494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4743931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4745357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4746805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4748243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4749677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4751101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4752567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4754004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4755428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4756863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4758358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4759881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4761284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4762736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4764166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4765613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4767023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4768459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4769882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4771326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4772765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4774182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4775629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4777109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4778656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4780123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4781583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4783055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4784499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4785920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4787368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4788817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4790252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4791682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4793151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4794619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4796083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4797605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4799088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4800547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4801972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4803428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4805312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4806795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4808218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4809653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4811090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4812555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4813970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4815482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4817037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4818541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4819981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4821413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4822874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4824310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4825745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4827177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4828617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4830044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4831476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4832909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4834408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4835936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4837415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4838832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4840259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4841708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4843139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4844575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4845999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4847439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4848852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4850289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4851712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4853234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4854736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4856209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4857633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4859066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4860494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4861909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4863364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4864794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4866209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4867615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4869051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4870469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4871932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4873456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4874937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4876364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4877790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4879201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4880628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4882080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4883519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4884949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4886386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4887851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4889271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4890748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4892265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4893808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4895239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4896685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4898108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4899547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4900959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4902397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4903839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4905559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4907002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4908418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4909930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4911480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4912992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4914414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4915858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4917288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4918706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4920120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4921563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4923013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4924434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4925870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4927294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4928828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4930429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4931927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4933368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4934811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4936218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4937638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4939060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4940498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4941907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4943346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4944793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4946219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4947684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4949185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4950672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4952103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4953558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4954980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4956413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4957847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4959281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4960697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4962140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4963597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4965013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4966486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4967990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4969470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4970874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4972298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4973743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4975177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4976587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4978011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4979428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4980867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4982280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4983713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4985171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4986694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4988166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4989581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4991022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4992458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4993939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4995360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4996815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4998245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.4999673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5001098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5002534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5004361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5006042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5007532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5008951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5010396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5011814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5013280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5014705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5016146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5017549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5018975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5020401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5021844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5023306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5024828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5026300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5027728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5029154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5030573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5032009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5033462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5034897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5036303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5037731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5039155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5040570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5042015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5043560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5045028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5046444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5047858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5049284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5050719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5052124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5053571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5054991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5056435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5057854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5059284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5060711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5062246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5063764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5065199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5066629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5068069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5069502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5070929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5072372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5073811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5075232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5076647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5078086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5079504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5081000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5082493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5083945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5085367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5086785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5088193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5089618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5091065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5092478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5093961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5095403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5096857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5098266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5099775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5101280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5102723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5104380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5105848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5107254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5108687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5110093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5111498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5112929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5114362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5121364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5122933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5125808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5128601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5131260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5133916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5136582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5139242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5141928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5144571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5147280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5149949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5152615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5155260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5157904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5160649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5163390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5166032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5168681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5171366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5174024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5176644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5179291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5181944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5184584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5187262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5189927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5192595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5195361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5198078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5200734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5203391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5206678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5209321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5211964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5214622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5217265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5219902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5222546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5225197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5227866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5230626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5233379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5236016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5238633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5241246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5243861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5246537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5249171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5251788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5254438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5257081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5259709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5262323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5265054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5267814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5270451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5273101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5275749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5278409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5281068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5283734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5286389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5289044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5291697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5294383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5297025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5299815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5302597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5305521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5308163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5310807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5313446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5316071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5318706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5321368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5324004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5326656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5329308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5331957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5334716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5337456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5340104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5342757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5345415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5348059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5350702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5353394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5356053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5358690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5361322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5363973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5366594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5369290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5372022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5374687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5377317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5379967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5382604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5385248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5387892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5390523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5393171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5395862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5398517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5401136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5403851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5407142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5409769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5412422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5415067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5417715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5420344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5422972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5425630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5428249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5430892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5433523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5436139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5438848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5441596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5444262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5446891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5449561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5452214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5454824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5457468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5460140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5462779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5465417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5468094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5470754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5473448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5476170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5478890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5481559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5484218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5486885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5489523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5492173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5494864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5497509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5500151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5502802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5505763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5508471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5511230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5513958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5516613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5519246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5521883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5524562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5527208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5529842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5532496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5535142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5537792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5540444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5543142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5545902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5549026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5551681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5554339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5556984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 48%] 2024-08-06T21:15:48.5559629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5562252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5564884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5567520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5570151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5572794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5575428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5578160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5580895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5583568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5586205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5588858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5591498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5594156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5596831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5599497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5602125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5605260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5607946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5610601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5613345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5616116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5618809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5621448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5624073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5626697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5629329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5631990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5634641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5637263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5639902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5642554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5645197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5647819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5650521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5653269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5655898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5658560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5661208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5663858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5666500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5669125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5671786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5674448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5677100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5679734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5682363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5685107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5687821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5690469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5693107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5695804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5698431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5701053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5703702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5706588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5709242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5711892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5714523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5717156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5719897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5722622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5725257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5727939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5730574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5733199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5735838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5738525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5741172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5743798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5746442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5749083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5751692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5754394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5757109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5759748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5762373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5765033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5767647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5770280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5772910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5775529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5778165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5780818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5783461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5786084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5788800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5791580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5794251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5796894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5799530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5802184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5805314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5807993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5810646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5813275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5815903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5818570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5821199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5823975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5826730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5829352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5831990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5834641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5837283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5839910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5842547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5845229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5847865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5850478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5853127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5855799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5858480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5861253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5863944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5866597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5869248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5871901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5874553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5877208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5879852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5882459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5885106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5887770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5890404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5893086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5895847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5898564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5901181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5903808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5906762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5909416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5912073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5914707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5917338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5920016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5922678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5925344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5928071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5930882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5933585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5936227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5938886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5941539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5944170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5946786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5949440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5952109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5954754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5957391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5960053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5962752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5965459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5968154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5970788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5973429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5976089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5978747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5981400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5984060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5986735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5989374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5992029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5994728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.5997377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6000070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6002796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6005937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6008573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6011190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6013836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6016470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6019114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6021745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6024401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6027051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6029691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6032298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6035064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6037840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6040509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6043185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6045850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6048504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6051170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6053819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6056480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6059160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6061840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6064494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6067150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6069896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6072627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6075260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6077910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6080600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6083234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6085897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6088550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6091214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6093888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6096575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6099247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6101907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6104946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6107721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6110375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6113047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6115687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6118355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6121036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6123719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6126390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6129041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6131678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6134340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6136996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6139671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6142406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6145112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6147793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6150417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6153075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6155751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6158401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6161041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6163690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6166335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6168987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6171616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6174344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6177098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6179797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6182417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6185073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6187737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6190390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6193045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6195738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6198359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6201017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6203633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6206784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6209569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6212342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6215020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6217664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6220327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6222979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6225610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6228287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6230950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6233602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6236275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6238947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6241635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6244294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6247054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6249793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6252477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6255158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6257826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6260481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6263162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6265803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6268443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6271098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6273766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6276421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6279057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6281777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6284539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6287194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6289869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6292530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6295220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6297862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6300507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6303171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6306109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6308779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6311435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6314059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6316856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6319635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6322294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6324950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6327603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6330221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6332842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6335487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6338153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6340809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6343473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6346123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6348764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6351492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6354211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6356858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6359545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6362213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6364836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6367485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6370185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6372848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6375494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6378192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6380856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6383506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6386197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6388941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6391639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6394314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6396975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6399616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6402272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6405462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6408134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6410767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6413403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6416070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6418698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6421422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6424277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6426985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6429599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6432278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6434974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6437625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6440266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6442920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6445583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6448220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6450874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6453542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6456195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6458931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6461650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6464274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6465709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6467109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6468551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6469955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6471389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6472784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6474195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6475600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6477029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6478440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6479904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6481408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6482820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6484236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6485641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6487065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6488501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6489915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6491313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6492720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6494125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6495576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6496960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6498479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6499967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6501355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6502766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6504451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6505894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6507279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6508717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6510131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6511559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6512957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6514376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6515783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6517322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6518854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6520257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6521691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6523110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6524523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6525923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6527342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6528781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6530186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6531576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6533002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6534408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6535859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6537357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6538834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6540264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6541656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6543081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6544504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6545946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6547368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6548816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6550247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6551705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6553117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6554592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6556101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6557597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6559034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6560458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6561895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6563311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6564730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6566142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6567573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6569014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6570443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6571854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6573334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6574843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6576320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6577732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6579192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6580626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6582035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6583472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6584912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6586363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6587780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6589227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6590657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6592142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6593644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6595168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6596589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6598040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6599457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6600867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6602303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6603735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6605640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6607074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6608526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6609954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6611477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6613016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6614516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6615938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6617361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6618794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6620229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6621660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6623074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6624502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6625917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6627361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6628796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 49%] 2024-08-06T21:15:48.6630226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6631712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6633231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6634629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6636052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6637473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6638929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6640338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6641772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6643180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6644599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6646026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6647434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6648899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6650400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6651910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6653322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6654771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6656212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6657641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6659082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6660533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6661965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6663386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6664830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6666249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6667685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6669210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6670744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6672169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6673611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6675024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6676452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6677872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6679337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6680749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6682184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6683618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6685049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6686473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6687960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6689498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6690932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6692364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6693788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6695266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6696700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6698138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6699576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6701011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6702435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6703860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6705567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6707112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6708681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6710096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6711532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6712959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6714404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6715814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6717251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6718680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6720120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6721531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6722962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6724380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6725877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6727390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6728820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6730267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6731696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6733123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6734537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6735970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6737390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6738816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6740223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6741657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6743074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6744623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6746129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6747540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6749003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6750412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6751838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6753258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6754716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6756133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6757561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6759008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6760466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6761877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6763392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6764907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6766357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6767775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6769225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6770663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6772082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6773515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6774929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6776366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6777793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6779250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6780663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6782170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6783674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6785094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6786510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6787941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6789404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6790804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6792234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6793653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6795131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6796555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6798019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6799451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6800964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6802454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6803879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6805787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6807234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6808634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6810042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6811477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6812900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6814317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6815721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6817154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6818566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6820087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6821626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6823122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6824544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6825969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6827377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6828805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6830252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6831667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6833096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6834517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6835964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6837374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6838854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6840352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6841830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6843228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6844650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6846065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6847498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6848921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6850336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6851764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6853182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6854605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6856009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6857483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6859010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6860484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6861892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6863328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6864751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6866172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6867587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6869053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6870476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6871883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6873319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6874723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6876201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6877688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6879175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6880588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6882022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6883430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6884851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6886261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6887701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6889122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6890529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6891962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6893380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6894892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6896394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6897878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6899316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6900738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6902149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6903581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6905269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6906713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6908117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6909565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6910978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6912377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6913789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6915299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6916839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6918234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6919674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6921090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6922519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6923916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6925339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6926761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6928198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6929628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6931042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6932478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6934575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6936102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6937511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6938950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6940400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6941820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6943230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6944662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6946077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6947492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6948896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6950348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6951765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6953241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6954763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6956178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6957616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6959025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6960468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6961887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6963334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6964740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6966162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6967582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6969033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6970442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6971926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6973446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6974874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6976304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6977722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6979163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6980581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6982001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6983405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6984836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6986254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6987683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6989098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6990601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6992103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6993506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6994971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6996399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6997841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.6999258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7000696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7002120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7003557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7005486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7006952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7008367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7009958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7011485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7012914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7014328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7015752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7017171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7018575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7020036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7021463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7022882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7024292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7025737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7027150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7028643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7030216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7031646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7033060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7034467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7035889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7037305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7038740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7040172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7041599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7043019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7044460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7045860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7047328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7048817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7050316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7051712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7053131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7054535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7055946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7057363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7058767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7060208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7061622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7063042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7064436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7065912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7067421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7068894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7070329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7071772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7073206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7074614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7076048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7077481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7078921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7080350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7081787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7083197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7084671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7086158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7087628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7089043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7090500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7091908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7093330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7094791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7096227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7097642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7099045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7100503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7101928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7103397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7105231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7106768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7108192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7109623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7111057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7112494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7113921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7115338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7116768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7118177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7119628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7121032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7122453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7123958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7125495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7126900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7128326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7129750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7131186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7132589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7134018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7135428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7136845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7138264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7139676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7141108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7142597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7144098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7145503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7146933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7148352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7149782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7151193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7152626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7154039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7155429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7156848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7158257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7159702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7161172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7162673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7164079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7165512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7166914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7168330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7169766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7171209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7172612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7174034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7175457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7176888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7178314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7179836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7181354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7182781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7184211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7185632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7187058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7188468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7189913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7191320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7192751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7194173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7195628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7197049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7198541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7200078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7201487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7202910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7204769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7206239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7207641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7209066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7210510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7211941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7213344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7214775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7216193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7217733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7219267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7220695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7222123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7223539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7224950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7226341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7227766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7229182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7230614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7232012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7233440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7234849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7236297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7237785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7239259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7240713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7242120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7243543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7244955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7246398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7247854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7249298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7250760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7252207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7253612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7255083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7256576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7258053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7259447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7260879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7262291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7263720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7265114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7266541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7267948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7269361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7270792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7272194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7273667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7275171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7276638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7278040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7279476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7280913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7282329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7283737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7285178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 50%] 2024-08-06T21:15:48.7286596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7287999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7289428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7290850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7292321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7293805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7295305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7296717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7298147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7299546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7300965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7302373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7303804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7305487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7306912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7308325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7309755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7311166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7312663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7314217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7315627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7317041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7318450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7319877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7321289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7322705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7324107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7325525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7326935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7328325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7329744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7331232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7332814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7334206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7335624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7337036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7338467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7339872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7341288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7342701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7344131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7345533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7346963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7348381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7349885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7352421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7354050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7355499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7356929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7358355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7359765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7361527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7364168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7365619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7367023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7368459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7370171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7372451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7374686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7376133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7379275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7381255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7382704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7384317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7388045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7391197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7394554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7396020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7398855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7400472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7403225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7406910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7408765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7411813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7413268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7415331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7416768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7418188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7420498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7421978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7423401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7425751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7427178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7429145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7431517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7433003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7434429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7435851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7437296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7438696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7440124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7441549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7442985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7444413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7445844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7447267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7448747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7450238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7451712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7453122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7454562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7455978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7457377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7458808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7460235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7461667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7463073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7464533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7465952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7467367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7468848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7470360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7471776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7473194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7474621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7476035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7477474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7478876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7480300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7481714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7483156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7484579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7485998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7487492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7489000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7490385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7491797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7493212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7494681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7496097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7497503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7498929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7500334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7501749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7503156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7504907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7506462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7507996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7509403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7510839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7512268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7513672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7515118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7516541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7517982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7519386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7520816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7522229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7523658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7525144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7526646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7528055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7529489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7530890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7532315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7533725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7535162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7536584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7537984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7539417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7540847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7542266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7543711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7545248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7546728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7548152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7549570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7551008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7552450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7553857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7555293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7556699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7558131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7559531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7560941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7562412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7563933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7565379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7566797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7568214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7569651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7571048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7572483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7573900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7575310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7576728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7578140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7579568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7581026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7582539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7583988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7585421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7586840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7588266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7589671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7591092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7592520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7593909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7595363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7596776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7598207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7599601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7601090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7602593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7604465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7606023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7607457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7608869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7610307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7611714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7613144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7614566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7616367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7617825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7619235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7620817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7622419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7623885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7625358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7626784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7628213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7629611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7631015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7632456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7633872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7635294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7636712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7638142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7639628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7641132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7642546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7643978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7645395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7646802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7648198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7649611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7651044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7652452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7653876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7655288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7656713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7658184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7659689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7661087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7662522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7663919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7665331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7666732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7668161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7669565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7670962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7672403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7673823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7675233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7676669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7678183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7679644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7681060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7682481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7683916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7685326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7686743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7688151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7689558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7690989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7692422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7693841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7695335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7696868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7698357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7699763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7701172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7702628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7704331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7705763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7707200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7708607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7710015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7711418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7712847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7714351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7715863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7717341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7718757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7720201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7721602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7723880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7725826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7727824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7729510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7731457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7733087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7735141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7737195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7738910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7741479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7744451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7747099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7749256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7752436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7754953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7757776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7759942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7761735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7763175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7764591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7766274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7768123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7769824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7781797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7783580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7785023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7786463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7787864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7789313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7790711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7792129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7793522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7794981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7796386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7797966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7799476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7800891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7802293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7803725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7805702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7807125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7808561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7810003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7811421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7812822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7814259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7815674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7817198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7818734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7820251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7821668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7823088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7824493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7825892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7827317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7828713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7830152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7831564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7832991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7834380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7835839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7837333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7838806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7840224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7841644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_143_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7843095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7844538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7845984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7847420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7848878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7850349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7851808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7853242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7854743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7856276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7857787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7859218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7860684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7862125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7863554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7864975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7866410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7867858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7869279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7870745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7872177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7873676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7875187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7876677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7878110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7879592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7881024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7882469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7883915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7885387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7886813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7888254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7889733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7891179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7892658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7894180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7895726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7897169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7898611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7900053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7901504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7902948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7904762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7906256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7907721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7909175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7910624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7912144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7913700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7915233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7916662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7918118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7919554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7921028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7922459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7923905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7925336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7926797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7928227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7929668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7931177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7932726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7934963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7936387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7937844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7939286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7940717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7942156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7943606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7945041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7946484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7947911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7949359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7950851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7952392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7953864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7955309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7956776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7958206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7959655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7961107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7962577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7964004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7965459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7966903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7968357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7969821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7971369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7973199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7974705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7976137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7977567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7979037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7980491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7981956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7983381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7984842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7986290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7987735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7989220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7990774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7992286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 51%] 2024-08-06T21:15:48.7993735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.7995219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.7996690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.7998133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.7999569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8001026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8002476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8004234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8005849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8007287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8008872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8010442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8011864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8013319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8014754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8016211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8017628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8019074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8020512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8021944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8023396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8024827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8026276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8027800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8029381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8030803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8032256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8033702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8035139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8036556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8038004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8039438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8040848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8042298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8043738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8045188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8046677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8048190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8049616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8051062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8052497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8053938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8055374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8056836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8058268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8059711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8061157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8062629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8064071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8065576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8067114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8068561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8070010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8071447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8072912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8074344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8075778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8077200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8078643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8080076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8081517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8082964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8084481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8086015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8087431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8088877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8090314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8091771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8093205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8094684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8096119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8097573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8098995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8100443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8101871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8103401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8105305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8106756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8108200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8109631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8111061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8112487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8113939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8115371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8116801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8118224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8119671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8121105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8122672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8124211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8125657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8127098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8128523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8129964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8131399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8132875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8134295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8135733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8137160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8138616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8140029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8141543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8143089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8144533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8145944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8147365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8148812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8150244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8151681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8153122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8154558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8155991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8157431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8158846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8160366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8161883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8163337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8164766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8166223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8167666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8169091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8170537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8171973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8173435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8174855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8176296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8177712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8179222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8180716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8182136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8183571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8185021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8186439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8187873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8189301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8190737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8192173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8193598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8195086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8196527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8198033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8199532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8200970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8202414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8203854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8205762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8207221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8208666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8210101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8211517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8212961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8214406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8215807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8217358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8218913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8220355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8221766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8223224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8224643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8226072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8227481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8228905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8230331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8231765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8233216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8234635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8236145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8237661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8239091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8240512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8241963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8243427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8244869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8246296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8247731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8249152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8250573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8251990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8253430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8254939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8256437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8257865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8259289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8260733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8262144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8263588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8265019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8266465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8267875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8269310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8270746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8272190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8273688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8275190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8276636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8278071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8279510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8280937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8282376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8283804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8285222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8286627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8288078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8289519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8290952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8292463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8293998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8295473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8296895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8298339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8299780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8301237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8302683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8304372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8305825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8307287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8308712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8310152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8311695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8313283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8314711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8316154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8317587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8319022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8320449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8321867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8323336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8324764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8326204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8327633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8329070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8330615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8332135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8333574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8335020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8336454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8337865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8339299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8340730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8342182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8343612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8345053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8346486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8347930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8349417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8350934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8352347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8353797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8355206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8356631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8358047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8359476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8360912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8362320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8363772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8365206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8366633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8368133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8369665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8371099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8372556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8373986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8375439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8376870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8378310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8379741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8381175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8382647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8384077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8385519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8387016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8388538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8389943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8391376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8392832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8394285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8395744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8397182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8398613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8400067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8401490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8402942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8404917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8406523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8408070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8409495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8410948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8412392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8413856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8415280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8416735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8418178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8419621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8421056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8422489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8423956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8425443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8426973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8428404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8429861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8431290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8432731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8434175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8435629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8437053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8438487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8439918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8441372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8442783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8444286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8445814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8447252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8448687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8450113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8451560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8452993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8454438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8455861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8457296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8458723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8460152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8461559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8463081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8464591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8466000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8467435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8468866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8470311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8471720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8473170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8474612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8476071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8477496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8478944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8480378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8481874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8483404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8484892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8486332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8487777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8489221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8490644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8492090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8493550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8495029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8496461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8497914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8499343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8500819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8502339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8503857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8505564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8507000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8508441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8509864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8511316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8512736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8514194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8515621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8517080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8518498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8519995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8521543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8523066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8524508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8525956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8527376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8528803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8530232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8531646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8533082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8534530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8535964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8537378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8538853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8540368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8541838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8543254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8544717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8546150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8547576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8548999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8550433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8551883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8553327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8554775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8556199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8557690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8559195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8560676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8562101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8563556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8564966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8566391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8567818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8569251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8570675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8572095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8573559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8574986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8576457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8577972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8579461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8580890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8582325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8583760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8585204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8586640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8588083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8589504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8590934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8592396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8593836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8595365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8596883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8598376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8599781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8601216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8602642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8604527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8605990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8607427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8608846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8610295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8611710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8613118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8614658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8616206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8617697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8619106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8620550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8621983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8623409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8624842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8626293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8627717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8629143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8630561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8631982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8633497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8635004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8636475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8637894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8639336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8640749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8642173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8643591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8645054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8646455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8647886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8649316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8650765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8652230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8653732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 52%] 2024-08-06T21:15:48.8655238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8656666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8658102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8659528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8660971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8662405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8663855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8665271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8666704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8668131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8669552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8671005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8672536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8674014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8675429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8676867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8678293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8679736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8681149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8682586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8684035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8685496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8686925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8688377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8689820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8691355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8692865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8694317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8695822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8697279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8698721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8700159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8701615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8703051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8704958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8706426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8707878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8709393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8710958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8712463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8713916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8715385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8716818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8718264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8719709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8721181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8722600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8724049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8725516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8726982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8728458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8730002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8731490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8732949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8734383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8735852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8737287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8738726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8740169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8741587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8743048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8744503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8745951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8747416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8748954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8750439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8751881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8753317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8754790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8756223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8757664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8759094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8760527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8761983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8763411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8764877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8766353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8767891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8769360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8770803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8772232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8773686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8775118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8776557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8777992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8779443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8780867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8782296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8783742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8785240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8786778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8788248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8789703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8791151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8792601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8794027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8795685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8797147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8798592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8800031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8801496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8802944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8804704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8806365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8807871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8809327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8810749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8812199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8813633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8815095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8816529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8817970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8819409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8820870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8822290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8823772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8825308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8826798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8828239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8829672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8831139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8832582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8834034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8835495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8836954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8838397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8839847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8841275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8842770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8844293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8845798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8847224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8848663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8850122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8851547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8852998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8854437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8855919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8857345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8858797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8860225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8861722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8863247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8864734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8866189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8867634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8869069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8870488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8871945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8873390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8874830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8876274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8877727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8879150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8880620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8882124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8883618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8885056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8886508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8887929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8889355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8890810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8892228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8893658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8895162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8896631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8898050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8899536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8901636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8903145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8905091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8906600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8908050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8909513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8910950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8912394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8913840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8915284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8916751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8918173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8919708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8921264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8922814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8924243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8925701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8927165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8928607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8930030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8931484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8932916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8934343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8935788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8937247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8938791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8940299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8941739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8943170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8944633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8946056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8947519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8948943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8950394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8951803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8953241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8954673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8956102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8957619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8959117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8960558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8961985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8963425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8964843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8966286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8967733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8969172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8970590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8972045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8973490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8974925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8976427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8977943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8979392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8980812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8982262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8983686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8985129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8986559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8988000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8989424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8990873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8992298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8993727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8995265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8996822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8998257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.8999677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9001130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9002563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9004195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9005646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9007128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9008561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9010003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9011429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9012874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9014415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9015966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9017411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9018838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9020286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9021697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9023125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9024551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9026007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9027437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9028875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9030303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9031752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9033240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9034757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9036182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9037650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9039073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9040493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9041939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9043378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9044824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9046242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9047709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9049152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9050586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9052047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9053568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9055033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9056461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9057880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9059323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9060751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9062164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9063600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9065020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9066478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9067898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9069330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9070798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9072351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9073816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9075259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9076714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9078174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9079594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9081023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9082476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9083909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9085352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9086803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9088241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9089706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9091223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9092686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9094132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9095604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9097066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9098486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9099930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9101368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9102782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9104662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9106172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9107649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9109150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9110707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9112213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9113673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9115103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9116547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9117998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9119457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9120885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9122330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9123758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9125192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9126616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9128091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9129623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9131100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9132539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9133962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9135409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9136840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9138285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9139707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9141154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9142593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9144034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9145450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9146960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9148533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9150003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9151448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9152892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9154349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9155765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9157226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9158659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9160107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9161520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9162959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9164387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9165860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9167397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9168867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9170313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9171750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9173183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9174601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9176041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9177499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9178928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9180348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9181800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9183233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9184705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9186214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9187708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9189164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9190588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9192030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9193448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9194896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9196350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9197802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9199233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9200686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9202095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9203567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9205325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9206842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9208276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9209699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9211152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9212593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9214034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9215462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9216919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9218378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9219822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9221244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9222764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9224342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9225830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9227256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9228719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9230156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9231564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9233001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9234433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9235884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9237326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9238773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9240200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9241691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9243199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9244686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9246117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9247584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9249008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9250426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9251874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9253319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9254754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9256180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9257657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9259091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9260573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9262083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9263575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9265004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9266440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9267882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9269326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9270759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9272185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9273618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9275049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9276505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9277945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9279423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9280933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9282441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9283856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9285293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9286721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9288193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9289616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9291057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9292481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9293919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9295395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9296825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9298321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9299831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9301301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9302716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9304607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9306082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9307507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9308940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9310389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9311813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9313246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9314664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9316092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 53%] 2024-08-06T21:15:48.9317633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9319186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9320686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9322128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9323595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9325020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9326462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9327921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9329383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9330802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9332248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9333674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9335101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9336570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9338090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9339580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9341013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9342451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9343871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9345314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9346749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9348203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9349619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9351059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9352488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9353914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9355367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9356891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9358406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9359824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9361263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9362698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9364149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9365566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9367008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9368446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9369889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9371300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9372721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9374177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9375701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9377153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9378589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9380030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9381467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9382887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9384301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9385748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9387171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9388626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9390047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9391487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9392957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9394478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9395981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9397426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9398880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9400296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9401732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9403152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9404815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9406274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9407701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9409143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9410588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9412062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9413610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9415089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9416530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9417939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9419372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9420810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9422243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9423676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9425090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9426539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9427984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9429422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9430881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9432428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9433915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9435352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9436775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9438233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9439653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9441066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9442501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9443923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9445363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9446776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9448234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9449695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9451222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9452676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9454105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9455526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9456974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9458398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9459831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9461260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9462687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9464118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9465543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9466990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9468482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9470004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9471463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9472893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9474317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9475741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9477139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9478597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9480025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9481426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9482851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9484274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9485714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9487118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9488640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9490150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9491598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9493010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9494449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9495909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9497365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9498803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9500246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9501670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9503107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9504996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9506452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9508029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9509584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9511011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9512425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9513871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9515293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9516719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9518137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9519593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9521016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9522449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9523860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9525280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9526847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9528346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9529791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9531215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9532672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9534082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9535519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9536947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9538407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9539826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9541266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9542687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9544110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9545607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9547097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9548544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9549972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9551403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9552810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9554246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9555681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9557107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9558523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9559971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9561406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9562834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9564321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9565845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9567294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9568726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9570169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9571595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9573046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9574467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9575908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9577326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9578789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9580197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9581620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9583112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9584617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9586041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9587457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9588926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9590353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9591777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9593195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9594633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9596098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9597531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9598970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9600408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9601902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9603414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9605038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9606478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9607932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9609366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9610797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9612207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9613645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9615038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9616463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9617882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9619339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9620802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9622344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9623820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9625239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9626660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9628063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9629523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9630957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9632392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9633801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9635248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9636688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9638114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9639589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9641137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9642613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9644047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9645469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9646885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9648320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9649766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9651199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9652619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9654063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9655476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9656906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9658374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9659922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9661377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9662797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9664224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9665659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9667086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9668505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9669965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9671393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9672826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9674254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9675697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9677173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9678691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9680154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9681587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9683012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9684441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9685848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9687275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9688741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9690153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9691587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9693014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9694455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9695946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9697465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9698938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9700381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9701792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9703223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9705112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9706599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9708012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9709429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9710868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9712297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9713724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9715133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9716695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9718227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9719650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9721056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9722493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9723907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9725324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9726734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9728145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9729597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9731003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9732428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9733849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9735368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9736869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9738299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9739742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9741197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9742606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9744040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9745469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9746916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9748326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9749775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9751212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9752629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9754124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9755614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9757051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9758471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9759926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9761336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9762772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9764198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9765628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9767041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9768460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9769929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9771332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9772826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9774331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9775773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9777182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9778616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9780057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9781500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9782913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9784344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9785749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9787185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9788589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9790014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9791530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9793028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9794442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9795885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9797325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9798743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9800174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9801582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9803018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9804651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9806095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9807513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9808932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9810462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9811995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9813478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9814903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9816355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9817766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9819188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9820617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9822055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9823444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9824869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9826289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9827721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9829209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9830725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9832203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9833619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9835037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9836448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9837883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9839304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9840745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9842151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9843593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9845017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9846438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9847894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9849404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9851469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9852881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9854315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9855730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9857163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9858562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9859984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9861403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9862839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9864243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9865666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9867120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9868639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9870098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9871506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9872938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9874359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9875770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9877176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9878620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9880064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9881490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9882901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9884333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9885746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9887238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9888730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9890145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9891572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9892975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9894387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9895835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9897278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9898677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9900107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9901521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9902951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9904828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9906410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9907943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9909380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9910800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9912219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9913653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9915076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9916505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9917924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9919362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9920800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9922229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9923636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9925134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9926644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9928060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9929469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9930906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9932340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9933740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9935165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9936576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9938009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9939411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9940860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9942292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9943821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9945317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9946757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9948195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9949656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9951108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9952537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9953994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9955445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9956883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9958314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9959765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9961214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9962719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9964213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9965656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9967091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9968533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9969954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9971418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9972853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9974280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 54%] 2024-08-06T21:15:48.9975716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9977149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9978609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9980023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9981556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9983072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9984526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9985947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9987394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9988829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9990283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9991722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9993147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9994583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9996053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9997572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:48.9998986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0000531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0002049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0003480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0005171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0006657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0008096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0009526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0010960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0012408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0013833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0015244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0016683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0018110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0019671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0021216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0022653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0024082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0025534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0026957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0028389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0029808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0031277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0032684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0034120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0035550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0036974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0038472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0039971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0041434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0042862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0044334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0045748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0047193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0048635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0050076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0051521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0052976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0054421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0055863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0057363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0058882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0060338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0061787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0063237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0064662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0066104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0067517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0068956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0070383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0071854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0073277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0074712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0076226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0077744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0079178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0080599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0082070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0083508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0084945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0086367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0087833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0089271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0090712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0092146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0093591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0095100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0096665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0098087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0099513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0100974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0102394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0103894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0105817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0107287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0108707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0110147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0111598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0113044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0114586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0116171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0117596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0119040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0120454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0121890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0123337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0124779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0127523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0130206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0132909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0135628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0138340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0141079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0143857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0146545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0149236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0151904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0154570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0157273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0159931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0162621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0165290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0167977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0170676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0173331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0176059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0178856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0181594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0184271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0186971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0189704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0192395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0195089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0197845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0200575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0203260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0206205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0208904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0211669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0214481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0217232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0219894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0222591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0225278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0227951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0230639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0233341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0236043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0238698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0241392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0244091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0246808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0249567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0252297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0254975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0257665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0260337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0263047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0265748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0268455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0271113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0273787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0276467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0279119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0281815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0284585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0287330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0290011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0292678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0295388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0298081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0300753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0303421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0306664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0309389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0312076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0314744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0317537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0320368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0323105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0325810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0328505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0331205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0333889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0336588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0339260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0341935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0344645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0347308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0349980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0352715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0355469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0358160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0360829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0363564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0366246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0368910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0371606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0374286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0376962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0379630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0382339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0385039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0387767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0390532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0393267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0396004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0398701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0401399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0404285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0406982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0409652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0412302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0414978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0417693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0420361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0423089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0425872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0428617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0431286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0433950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0436657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0439346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0442002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0444696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0447379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0450075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0452759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0455447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0458159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0460939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0463677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0466348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0469025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0471728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0474379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0477041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0479714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0482390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0485062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0487727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0490389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0493129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0495928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0498655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0501338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0510901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0513728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0516402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0519095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0521803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0524472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0527147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0529825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0532513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0535308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0538140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0540899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0543562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0546230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0548874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0551519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0554199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0556882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0559532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0562183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0564950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0567625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0570270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0573067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0575881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0578541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0581209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0583884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0586567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0589238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0591907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0594619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0597362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0600043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0602703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0605721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0608534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0611298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0613986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0616643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0619317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0621996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0624646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0627315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0629998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0632678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0635332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0638007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0640711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0643466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0646210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0648909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0651601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0654273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0656938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0659603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0662285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0664957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0667657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0670314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0672979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0675620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0678335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0681127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0683791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0686469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0689120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0691772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0694462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0697164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0699806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0702453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0705402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0708072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0710706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0713475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0716249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0718902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0721561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0724241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0726911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0729576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0732220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0734861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0737499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0740154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0742815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0745440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0748211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0750947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0753569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0756218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0758883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0761537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0764167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0766835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0769502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0772150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0774812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0777475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0780169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0782910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0785643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0788294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0790977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0793670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0796368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0799048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0801708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0804583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0807240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0809892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0812557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0815203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0817970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0820759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0823420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0826071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0828719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0831383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0834056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0836722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0839376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0842034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0844711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0847399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0850048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0852803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0855557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0858221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0860879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0863547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0866209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0868851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0871494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0874167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0876828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0879484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0882141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0884782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0887515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0890264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0892919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0895601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0898270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0900933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0903567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0906535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0909226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0911877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0914527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0917181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0919861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0922611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0925368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0928044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0930694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0933336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0935958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0938584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0941239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0943884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0946562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0949212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0951874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0954514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0957184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0959934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0962640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0965324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0967978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0970634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0973326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0975997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0978659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0981325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0984031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0986688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0989336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0992029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0994813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.0997571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1000256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1002957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1005899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1008571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1011240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1013926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1016612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1019304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1021978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1024643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1027413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1030204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1032923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1035598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1038271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1040960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1043622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1046313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1049004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1051680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1054340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1057003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1059668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1062360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1065085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1067795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1070464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1073130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1075801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 55%] 2024-08-06T21:15:49.1078465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1081159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1083823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1086471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1089122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1091810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1094483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1097239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1100029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1102749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1105632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1108307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1110977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1113679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1116354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1119025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1121674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1124344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1127037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1129682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1132436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1135243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1137965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1140603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1143280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1145964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1148623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1151294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1153984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1156648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1159317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1161968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1164643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1167322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1170090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1172812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1175489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1178178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1180878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1183536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1186204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1188883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1191544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1194178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1196881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1199561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1202214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1205216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1208063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1210737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1213394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1216054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1218713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1221380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1224040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1226707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1229374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1232057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1234755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1237415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1240187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1242953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1245640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1248300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1250966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1253621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1256252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1258887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1261565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1264249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1266896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1269554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1272185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1274928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1277653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1280288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1282969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1285655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1288330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1290981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1293661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1296387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1299049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1301731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1304638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1307320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1310108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1312876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1315557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1318224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1320891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1323525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1326180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1328853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1331505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1334146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1336805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1339468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1342151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1344875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1347640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1350321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1352975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1355649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1358324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1361004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1363680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1366358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1369060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1371749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1374428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1377101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1379844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1382604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1385260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1387915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1390574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1393282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1396014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1398677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1401350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1404242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1406915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1409593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1412284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1415033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1418522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1421244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1423931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1426630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1429317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1432004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1434698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1437399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1440058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1442738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1445409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1448076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1450823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1453565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1456276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1458944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1461618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1464284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1466949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1469663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1472339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1474977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1477672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1480364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1483026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1484479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1486044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1487512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1488935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1490351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1491792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1493215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1494640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1496123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1497556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1498970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1500371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1501793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1503248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1505005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1506555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1507983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1509398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1510845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1512239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1513664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1515089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1516555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1517967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1519401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1520829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1522257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1523772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1525297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1526756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1528186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1529628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1531041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1532479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1533901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1535318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1536735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1538177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1539598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1541003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1542500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1543998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1545438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1546858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1548290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1549708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1551153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1552569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1554001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1555433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1556885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1558297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1559728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1561247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1562760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1564185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1565612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1567052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1568470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1569888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1571303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1572742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1574164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1575602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1577014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1578442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1579933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1581444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1582852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1584269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1585729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1587143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1588569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1589995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1591436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1592838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1594263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1595744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1597188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1598663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1600171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1601577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1602996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1604628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1606071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1607509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1608931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1610348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1611747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1613181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1614599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1616032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1617537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1619108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1620533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1621963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1623382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1624812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1626277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1627696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1629142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1630566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1632021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1633442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1634873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1636382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1637903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1639298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1640723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1642149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1643572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1644989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1646423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1647869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1649294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1650718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1652136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1653570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1655068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1656596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1658048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1659484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1660907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1662330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1663737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1665156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1666617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1668021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1669455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1670871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1672300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1673737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1675235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1676708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1678143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1679541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1680965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1682372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1683787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1685206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1686624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1688057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1689478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1690899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1692342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1693879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1695346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1696825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1698242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1699683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1701101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1702530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1704134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1705568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1707013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1708414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1709833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1711244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1712783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1714293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1715711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1717135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1718576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1719972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1721395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1722814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1724235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1725658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1727081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1728521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1729940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1731439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1732931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1734365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1735801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1737228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1738636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1740065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1741487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1742900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1744302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1745730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1747168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1748565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1750058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1751549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1752983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1754384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1755817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1757229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1758661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1760066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1761486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1762901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1764323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1765760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1767168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1768691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1770193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1771612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1773016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1774442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1775867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1777274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1778668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1780098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1781506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1782922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1784322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1785749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1787222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1788702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1790164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1791579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1793024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1794430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1795913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1797343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1798788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1800189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1801619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1803041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1804687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1806252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1807787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1809271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1810689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1812104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1813508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1814936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1816374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1817799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1819199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1820631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1822057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1823470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1824913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_2048_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1826448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1827943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1829361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1830800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1832243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1833696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1835117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1836591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1838020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1839476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1840902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1842343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1843820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1845335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1846828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1848246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1849702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1851140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1852574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1853997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1855450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1856898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1858331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1859751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1861192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1862666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1864182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1865648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1867095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1868551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1869977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1871421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1872860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1874321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1875761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1877211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1878640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1880431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1881938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1883466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1884934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1886394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1887817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1889242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1890685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1892120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1893559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1894977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1896486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1897927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1899353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1900810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1902340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1903829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1905561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1907011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1908468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1909901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1911322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1912764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1914182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1915622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1917057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1918489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1919984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1921552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 56%] 2024-08-06T21:15:49.1923026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1924455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1925877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1927348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1928752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1930170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1931623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1933061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1934503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1935932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1937410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1938896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1940423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1941944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1943399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1944838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1946284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1947724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1949169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1950604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1952017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1953455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1954889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1956341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1957815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1959339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1960809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1962264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1963685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1965128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1966571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1968032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1969461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1970901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1972337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1973785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1975224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1976704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1978258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1979751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1981195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1982626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1984074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1985499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1986952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1988381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1989824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1991256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1992700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1994127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1995602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1997208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.1998677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2000113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2001544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2002998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2004624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2006093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2007555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2009008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2010426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2011867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2013298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2014796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2016344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2017846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2019289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2020718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2022152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2023561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2025004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2026435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2027879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2029292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2030733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2032161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2033626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2035125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2036604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2038071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2039492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2040941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2042373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2043836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2045259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2046700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2048146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2049607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2051031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2052514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2054045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2055536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2056961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2058381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2059837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2061271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2062705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2064133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2065575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2067015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2068458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2069881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2071368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2072886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2074362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2075776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2077231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2078672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2080087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2081527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2082965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2084419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2085846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2087296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2088720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2090196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2091687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2093154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2094571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2096080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2097502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2098908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2100348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2101781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2103206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2104870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2106392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2107835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2109266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2110795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2112354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2113787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2115224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2116677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2118122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2119561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2120980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2122417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2123841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2125297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2126730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2128161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2129672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2131200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2132604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2134035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2135459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2136923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2138336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2139771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2141190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2142619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2144057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2145473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2146943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2148453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2149974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2151392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2152846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2154287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2155717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2157155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2158592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2160015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2161418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2162849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2164273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2165715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2167210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2168720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2170135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2171582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2172994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2174421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2175847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2177296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2178709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2180379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2181840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2183275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2184697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2186206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2187755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2189182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2190616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2192037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2193469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2194884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2196359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2197775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2199210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2200638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2202067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2203476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2205300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2206894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2208304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2209737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2211171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2212614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2214029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2215464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2216915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2218365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2219782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2221229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2222654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2224164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2225682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2227123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2228566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2229996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2231414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2232825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2234268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2235695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2237144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2238560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2240004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2241425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2242922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2244460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2245882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2247358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2248779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2250211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2251639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2253097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2254506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2255934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2257381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2258838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2260247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2261752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2263265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2264702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2266110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2267542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2268981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2270410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2271842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2273259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2274693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2276119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2277565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2278974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2280494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2282001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2283431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2284846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2286277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2287726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2289137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2290576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2292004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2293449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2294863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2296359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2297784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2299303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2300796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2302216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2303638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2305344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2306783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2308197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2309638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2311063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2312482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2313892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2315334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2316771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2318297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2319810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2321246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2322671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2324102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2325514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2326972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2328400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2329817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2331240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2332651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2334090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2335483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2336972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2338490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2339965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2341367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2342794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2344204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2345637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2347063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2348486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2349927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2351359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2352792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2354201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2355690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2357233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2358709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2360129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2361579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2363014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2364446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2365868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2367324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2368750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2370159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2371592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2373012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2374490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2375987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2377481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2378908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2380354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2381760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2383188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2384615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2386067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2387482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2388916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2390347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2391783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2393257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2394762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2396553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2398000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2399445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2400863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2402291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2403718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2405395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2406835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2408286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2409718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2411126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2412636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2414175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2415679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2417100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2418542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2419963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2421396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2422805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2424235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2425647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2427110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2428529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2429953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2431413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2432927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2434398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2435808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2437263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2438683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2440095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2441499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2442943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2444358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2445775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2447209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2448641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2450099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2451590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2453056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2454477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2455927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2457367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2458797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2460228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2461684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2463092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2464527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2465958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2467433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2468888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2470408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2471885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2473306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2474734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2476149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2477607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2479033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2480466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2481878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2483313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2484741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2486176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2487609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2489121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2490627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2492045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2493460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2494888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2496373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2497809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2499247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2500664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2502105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2503520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2505207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2506644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2508199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2509704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2511122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2512540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2513967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2515389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2516795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2518237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2519654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2521075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2522486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2523920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2525336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2526857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2528350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2529784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2531207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2532640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2534049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2535471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2536931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2538342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2539777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2541193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2542629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2544025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2545562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2547090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2548528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2549932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2551364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2552777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2554210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2555618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2557044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2558478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2559900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2561327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2562731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2564241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2565746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2567193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2568611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2570051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2571474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2572901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2574323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2575731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2577180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2578578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 57%] 2024-08-06T21:15:49.2579999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2581408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2582913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2584394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2585816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2587249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2588692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2590094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2591516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2592938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2594377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2595780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2597258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2598703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2600122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2601617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2603107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2604758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2606203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2607655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2609067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2610492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2611903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2613321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2614717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2616132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2617595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2618990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2620478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2622019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2623511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2624908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2626335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2627778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2629214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2630623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2632054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2633472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2634921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2636339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2637767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2639245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2640756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2642220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2643637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2645073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2646487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2647916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2649322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2650756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2652170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2653595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2655010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2656420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2657913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2659401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2660869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2662302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2663763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2665186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2666640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2668080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2669541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2670957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2672408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2673848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2675299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2676780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2678299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2679785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2681211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2682643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2684066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2685512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2686974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2688420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2689837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2691287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2692725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2694156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2695630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2697234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2698724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2700144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2701587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2703035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2704703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2706198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2707684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2709118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2710578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2712010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2713453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2714945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2716506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2718010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2719441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2720878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2722321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2723761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2725187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2726640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2728090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2729527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2730954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2732394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2733862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2735376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2736837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2738301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2739745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2741169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2742606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2744037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2745492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2746908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2748354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2749777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2751216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2752662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2754177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2755657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2757128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2758543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2759985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2761404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2762838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2764268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2765683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2767141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2768583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2770022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2771484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2773024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2774507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2775945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2777393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2778853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2780292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2781736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2783168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2784589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2786035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2787476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2788910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2790379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2791919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2793384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2794817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2796293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2797777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2799191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2800630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2802068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2803520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2805146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2806601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2808078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2809581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2811139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2812623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2814072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2815514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2816959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2818399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2819844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2821276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2822703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2824117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2825553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2827004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2828479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2830021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2831492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2832942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2834363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2835806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2837226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2838679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2840099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2841533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2842959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2844413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2845836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2847351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2848884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2850365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2851796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2853214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2854656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2856073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2857507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2858923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2860360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2861784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2863214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2864631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2866113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2867642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2869102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2870536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2871975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2873443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2874861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2876309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2877775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2879232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2880656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2882106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2883543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2885042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2886552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2888063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2889490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2890925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2892364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2893780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2895229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2896707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2898170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2899593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2901042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2902474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2904187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2905760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2907271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2908720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2910138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2911580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2913006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2914459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2915882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2917320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2918765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2920219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2921637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2923114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2924616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2926099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2927504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2928948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2930381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2931804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2933232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2934649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2936083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2937526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2938960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2940372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2941855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2943369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2944841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2946254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2947711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2949155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2950595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2952017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2953446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2954896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2956310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2957769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2959194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2960639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2962112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2963614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2965032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2966479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2967918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2969345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2970768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2972199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2973621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2975039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2976485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2977939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2979369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2980872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2982398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2983832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2985268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2986691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2988162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2989599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2991043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2992460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2993905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2995332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2996778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2998232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.2999736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3001260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3002673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3004325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3005819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3007269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3008700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3010135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3011554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3012992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3014404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3015817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3017254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3018811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3020345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3021759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3023205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3024637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3026065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3027479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3028920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3030336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3031743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3033155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3034590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3036008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3037483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3039010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3040424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3041863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3043273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3044696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3046116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3047571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3048984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3050411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3051837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3053291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3054702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3056195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3057723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3059155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3060586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3062012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3063441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3064863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3066288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3067711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3069145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3070569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3072000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3073407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3074909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3076410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3077833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3079259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3080684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3082121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3083528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3084955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3086380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3087846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3089267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3090702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3092122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3093640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3095136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3096602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3098047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3099477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3100898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3102306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3103744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3105419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3106842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3108280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3109719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3111135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3112658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3114190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3115622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3117045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3118494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3119911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3121335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3122789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3124203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3125629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3127053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3128525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3129934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3131428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3132974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3134405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3135808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3137236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3138671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3140092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3141517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3142929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3144363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3145788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3147214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3148622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3150120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3151617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3153027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3154431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3155868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3157299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3158718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3160132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3161551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3162990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3164396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3165823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3167238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3168738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3170208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3171623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3173031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3174462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3175860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3177286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3178700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3180122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3181532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3182936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3184380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3185799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3187280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3188797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3190278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3191704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3193137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3194553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3195987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3197480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3198915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3200329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3201748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3203186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3204800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3206300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3207859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3209354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3210758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3212184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3213600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3215032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3216445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3217897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3219312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3220754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3222173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3223584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3225063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3226576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3228065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3229482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3230925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3232351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3233768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3235184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 58%] 2024-08-06T21:15:49.3236612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3238047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3239468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3240880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3242296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3243770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3245261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3246730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3248165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3249609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3251023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3252448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3253864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3255304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3256704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3258141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3259567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3261002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3262464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3263965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3265443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3266883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3268308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3269719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3271144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3272556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3273963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3275354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3276778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3278197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3279612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3281057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3282575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3284033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3285431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3286869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3288300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3289735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3291143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3292577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3293998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3295444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3296915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3298361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3299824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3301358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3302811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3304429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3313116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3314717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3316138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3317552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3318967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3320380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3321786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3323193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3324634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3326185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3327735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3329223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3330634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3332050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3333459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3334883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3336298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3337714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3339117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3340522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3341939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3343358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3344776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3346259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3347739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3349149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3350537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3351938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3353337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3354773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3356165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3357558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3358957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3360373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3361770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3363159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3365029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3366556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3367954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3369351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3370768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3372175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3373576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3374994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3376408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3377816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3379226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3380628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3382023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3383503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3384982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3386374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3387781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3389198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3390589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3391988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3393394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3394810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3396255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3397674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3399088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3400500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3401972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3403492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3405207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3406652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3408063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3409459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3410872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3412289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3413705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3415106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3416511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3417920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3419307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3420831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3422355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3423789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3425191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3426598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3428000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3429410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3430806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3432202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3433619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3435035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3436428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3437826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3439305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3440856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3442248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3443668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3445085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3446496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3447894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3449294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3450705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3452103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3453516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3454906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3456311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3457790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3459268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3460659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3462065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3463510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3464899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3466292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3467705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3469130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3470521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3471930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3473370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3474784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3476218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3477726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3479181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3480608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3482015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3483443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3484847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3486258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3487655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3489049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3490464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3491874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3493287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3494726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3496256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3497720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3499115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3500511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3501918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3503349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3505007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3506441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3507850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3509268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3510674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3512082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3513577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3515122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3516599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3518004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3519406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3520822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3522209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3523623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3527257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3532315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3535507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3536912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3538357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3539793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3541207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3542604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3544012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3545426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3546817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3548216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3549634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3551060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3552596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3554109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3555588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3557007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3558416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3559811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3561214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3562619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3564014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3565407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3566820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3568233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3569616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3571077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3572588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3574057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3575444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3576845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3578243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3579674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3581061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3582451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3583846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3585256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3586647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3588039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3589506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3591006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3592457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3593856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3595261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3596705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3598100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3599490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3600892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3602285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3603671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3605395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3606827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3608313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3609831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3611311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3612717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3614141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3615532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3616934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3618347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3619780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3621178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3622587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3623999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3625412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3626848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3628338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3629807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3631211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3632606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3633997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3635400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3636808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3638204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3639614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3641020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3642426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3643823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3645258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3646762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3648216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3649631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3651042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3652453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3653876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3655269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3656681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3658083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3659526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3660929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3662329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3663766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3665263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3666696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3668086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3669518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3670933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3672330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3673725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3675136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3676544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3677937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3679357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3680766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3682158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3683600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3685094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3686554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3687964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3689382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3690785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3692192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3693606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3694996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3696435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3697839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3699263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3700632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3702069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3703550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3705283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3706693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3708083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3709518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3710917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3712312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3713696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3715106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3716519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3717924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3719339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3720827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3722360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3723826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3725231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3726641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3728062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3729474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3730887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3732288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3733695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3735084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3736486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3737889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3739374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3740892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3742347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3743745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3745164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3746549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3747937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3749355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3750758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3752149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3753536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3754953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3756357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3757794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3759307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3760770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3762179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3763586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3764981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3766370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3767771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3769169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3770557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3771947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3773351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3774730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3776167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3777643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3779118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3780497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3781883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3783285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3784695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3786081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3787470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3788897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3790301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3791696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3793087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3794537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3796032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3797519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3798925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3800326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3801724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3803109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3804705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3806127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3807556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3808957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3810364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3811767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3813182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3814638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3816162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3817626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3819059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3820454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3821857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3823261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3824685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3826080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3827479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3828913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3830419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3831824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3833277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3834775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3836219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3837610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3838997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3840398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3841801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3843197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3844587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3845978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3847424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3848816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3850214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3851664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3853183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3854623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3856018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3857429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3858851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3860234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3861639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3863043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3864448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3865860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3867272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3868679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3870117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3871600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3873028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3874440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3875840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3877238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3878619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3880024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3881426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3882820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3884207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3885609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3887043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3888474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3889967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3891432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3892858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3894253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3895673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3897124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3898560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 59%] 2024-08-06T21:15:49.3899960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3901374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3902772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3904414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3905885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3907288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3908768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3910371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3911841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3913234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3914652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3916057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3917463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3918861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3920290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3921713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3923119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3924537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3925958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3927452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3928945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3930417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3931844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3933297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3934708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3936123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3937551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3939069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3940460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3941875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3943289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3944702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3946156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3947667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3949142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3950564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3951989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3953396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3954828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3956254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3957701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3959111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3960549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3961975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3963401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3964852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3966377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3967887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3969299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3970726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3972137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3973565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3974960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3976371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3977806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3979243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3980645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3982063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3983517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3985017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3986476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3987908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3989338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3990751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3992163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3993566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3994994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3996446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3997886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.3999299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4000731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4002196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4003699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4005372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4006785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4008223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4009612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4011021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4012433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4013860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4015254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4016669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4018085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4019505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4020963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4022491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4023964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4025404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4026821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4028231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4029666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4031096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4032523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4033935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4035372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4036816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4038244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4039743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4041274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4042733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4044146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4045554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4046987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4048420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4049830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4051255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4052666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4054102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4055502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4056943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4058395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4059909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4061354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4062777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4064203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4065652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4067087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4068509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4069950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4071375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4072797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4074213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4075638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4077118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4078617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4080063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4081497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4082923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4084342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4085743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4087198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4088620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4090025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4091450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4092866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4094289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4095723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4097280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4098747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4100170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4101571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4102987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4104611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4106067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4107496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4108899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4110321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4111732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4113144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4114622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4116163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4117656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4119075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4120480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4121908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4123312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4124715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4126112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4127550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4128968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4130375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4131803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4133267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4134798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4136262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4137702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4139125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4140569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4141974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4143398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4144811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4146253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4147650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4149068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4150480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4151946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4153453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4154910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4156359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4157785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4159198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4160596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4162019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4163435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4164849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4166272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4167713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4169134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4170575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4172075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4173546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4174977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4176404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4177831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4179350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4180779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4182177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4183595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4185005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4186464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4187867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4189324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4190820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4192280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4193690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4195089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4196584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4198014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4199426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4200832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4202253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4203667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4205304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4206782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4208279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4209815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4211291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4212709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4214113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4215542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4216959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4218375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4219788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4221222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4222622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4224039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4225449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4226895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4228336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4229827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4231283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4232707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4234124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4235527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4236968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4238389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4239803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4241212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4242644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4244069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4245478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4246934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4248443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4249894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4251300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4252698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4254104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4255539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4256946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4258365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4259775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4261212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4262610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4264027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4265485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4266993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4268427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4269847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4271267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4272680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4274110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4275518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4276941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4278349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4279761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4281163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4282576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4284044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4285536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4286974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4288390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4289800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4291196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4292591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4294015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4295441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4296867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4298285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4299702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4301136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4302579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4304365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4306390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4307841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4309256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4310687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4312102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4313529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4314960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4316365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4317795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4319213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4320622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4322074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4323630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4325129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4326542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4327951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4329378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4330788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4332207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4333628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4335047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4336480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4337886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4339307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4340769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4342285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4343745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4345177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4346597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4348036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4349438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4350866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4352281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4353715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4355133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4356538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4357965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4359421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4360921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4362371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4363822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4365241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4366656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4368055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4369488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4370908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4372329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4373756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4375177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4376616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4378063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4379573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4381039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4382481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4383914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4385342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4386754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4388184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4389580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4390998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4392410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4393870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4395270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4396764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4398276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4399738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4401155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4402562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4404237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4405687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4407105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4408509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4409941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4411355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4412778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4414208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4415684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4417230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4418683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4420106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4421508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4422937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4424344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4425755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4427162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4428799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4430200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4431621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4433028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4434470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4435903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4437386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4438869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4440291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4441711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4443120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4444573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4446001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4447422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4448829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4450267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4451691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4453109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4454577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4456076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4457566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4458961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4460384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4461800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4463240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4464645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4466071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4467482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4468920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4470317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4471735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4473196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4474718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4476159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4477560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4478994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4480419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4481842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4483266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4484704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4486119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4487542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4488955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4490380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4491831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4493341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4494785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4496199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4497671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4499078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4500501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4501915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4503370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4505112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4506586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4508001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4509431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4510884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4512415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4513907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4515344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4516748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4518152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4519581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4520996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4522413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4523837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4525256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4526668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4528076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4529509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4531026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4532480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4533923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4535324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4536731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4538153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4539545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4540958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4542373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4543830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4545236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4546655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4548110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4549632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4551078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 60%] 2024-08-06T21:15:49.4552508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4553953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4555392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4556790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4558201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4559629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4561043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4562454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4563881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4565308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4566765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4568267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4569716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4571146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4572564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4573999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4575403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4576811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4578245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4579640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4581056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4582471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4583927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4585364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4586871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4588323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4589748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4591143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4592566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4593977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4595397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4596826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4598226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4599645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4601056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4602465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4604138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4605729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4607204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4608611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4610006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4611432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4612843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4614264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4615664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4617075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4618508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4619913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4621339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4622748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4624233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4625763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4627238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4628641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4630067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4631454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4632871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4634281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4635710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4637100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4638502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4639922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4641329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4642789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4644278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4645750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4647159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4648571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4649967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4651392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4652822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4654238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4655637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4657052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4658489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4659886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4661345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4662847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4664318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4665710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4667122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4668525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4669950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4671344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4672778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4674181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4675605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4677002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4678398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4679870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4681362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4682837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4684234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4685665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4687081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4688494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4689903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4691328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4692757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4694170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4695572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4697015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4698482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4699952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4701399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4702823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4704463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4705869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4707276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4708676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4710103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4711491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4712918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4714330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4715764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4717221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4718738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4720229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4721643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4723091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4724498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4725924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4727337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4728757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4730157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4731570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4733001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4734408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4735841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4737332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4738815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4740205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4741618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4743049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4744478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4745867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4747283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4748692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4750122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4751520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4752959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4754443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4755993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4757454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4758856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4760286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4761706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4763148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4764554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4765978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4767388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4768797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4770196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4771624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4773055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4774511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4775993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4777448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4778876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4780273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4781691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4783126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4784561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4785956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4787366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4788782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4790214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4791609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4793091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4794655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4796141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4797581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4798999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4800422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4801831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4803252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4804862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4806334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4807761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4809183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4810587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4812077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4813633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4815112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4816513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4817922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4819347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4820738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4822156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4823581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4825005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4826399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4827822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4829231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4830696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4832173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4833650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4835048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4836473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4837858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4839243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4840660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4842071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4843495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4844887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4846309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4847718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4849161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4850638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4852115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4853544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4854957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4856359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4857770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4859204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4860606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4862040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4863458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4864901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4866302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4867764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4869256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4870738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4872145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4873557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4874971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4876400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4877793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4879195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4880619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4882041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4883459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4884860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4886320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4887835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4889295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4890694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4892138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4893561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4894978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4896380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4897839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4899275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4900678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4902118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4903525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4905276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4906751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4908215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4909680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4911108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4912530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4913949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4915354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4916787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4918182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4919583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4921008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4922437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4923845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4925279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4926836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4928299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4929713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4931118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4932568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4933985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4935399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4936805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4938203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4939628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4941012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4942443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4943888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4945399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4946843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4948254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4949659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4951084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4952490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4953909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4955324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4956739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4958153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4959565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4960998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4962490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4964005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4965453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4966893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4968313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4969732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4971140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4972589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4973999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4975403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4976804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4978216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4979644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4981076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4982599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4984054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4985482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4986879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4988300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4989699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4991120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4992539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4993951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4995357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4996807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4998221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.4999657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5001164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5002651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5004294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5005718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5007137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5008538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5009941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5011329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5012772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5014180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5015593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5016988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5018453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5019993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5021459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5022888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5024294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5025726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5027117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5028533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5029944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5031373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5032786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5034211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5035625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5037030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5038483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5039974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5041435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5042866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5044274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5045664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5047077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5048486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5049896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5051287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5052731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5054140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5055526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5056992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5058493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5059967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5061362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5062804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5064214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5065644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5067045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5068463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5069864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5071292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5072707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5074124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5075561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5077057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5078506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5079898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5081322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5082749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5084157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5085552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5086969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5088373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5089784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5091178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5092614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5094063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5095545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5097039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5098452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5099888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5101279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5102706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5104337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5105824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5107219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5108633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5110025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5111440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5112900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5114419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5115875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5117272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5118672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5120063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5121473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5122888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5124291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5125683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5127105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5128523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5129932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5131364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5132888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5134345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5135737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5137154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5138563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5139994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5141393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5142829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5144226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5145644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5147033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5148439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5149836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5151305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5152805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5155541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5158189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5160856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5163488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5166131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5168808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5171584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5174263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5176938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5179623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5182317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5185038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5187827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5190556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5193248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5195943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5198699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5201377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5204287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5206980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5209610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5212280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5214980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5217646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5220368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5223170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5225906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5228577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5231238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5233910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5236596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5239272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5241952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5244625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 61%] 2024-08-06T21:15:49.5247312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5250007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5252674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5255394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5258170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5260868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5263553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5266226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5268919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5271573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5274225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5276875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5279543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5282227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5284889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5287556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5290318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5293054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5295783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5298486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5301184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5303842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5306729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5309385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5312071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5314738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5317394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5320095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5322804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5325548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5328309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5331028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5333693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5336346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5339018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5341671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5344317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5346979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5349646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5352301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5354973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5357662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5360337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5363600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5366332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5369001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5371668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5374345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5377058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5379732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5382401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5385098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5387785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5390463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5393107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5395831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5398633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5401336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5404221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5406942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5409633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5412286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5414958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5417633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5420305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5422964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5425621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5428272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5431047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5433844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5436585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5439264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5441955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5444619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5447286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5449965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5452674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5455352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5458042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5460699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5463378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5466075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5468829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5471542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5474219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5476884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5479531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5482198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5484894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5487569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5490224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5492894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5495550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5498244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5500890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5503596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5506664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5509392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5512050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5514720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5517393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5520059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5522716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5525385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5528032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5530682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5533318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5535961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5538715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5541474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5544162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5546817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5549484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5552133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5554770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5557440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5560125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5562799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5565497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5568171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5570862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5573592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5576357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5579062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5581777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5584452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5587120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5589789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5592488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5595127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5597809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5600499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5603191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5606090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5608813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5611575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5614311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5616969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5619671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5622336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5625007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5627648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5630293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5632950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5635621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5638312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5640969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5643674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5646441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5649150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5651797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5654462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5657143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5659789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5662407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5665059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5667717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5670364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5673037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5675714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5678402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5681157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5683848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5686505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5689187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5691851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5694507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5697210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5699922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5702591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5705498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5708186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5710854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5713596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5716379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5719093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5721751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5724405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5727066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5729698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5732381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5735036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5737674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5740327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5742989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5745617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5748305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5751056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5753796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5756443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5759088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5761738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5764411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5767072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5769739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5772403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5775077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5777722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5780400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5783095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5785833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5788535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5791167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5793800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5796462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5799156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5801803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5804688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5807403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5810059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5812683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5815344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5817998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5820720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5823498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5826255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5828907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5831568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5834244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5836896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5839569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5842247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5844873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5847523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5850169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5852798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5855477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5858206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5860904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5863547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5866181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5868853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5871510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5874153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5876770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5879426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5882096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5884760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5887404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5890108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5892861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5895604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5898309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5900983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5903663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5906567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5909230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5911869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5914538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5917203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5919850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5922525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5925254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5928012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5930716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5933368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5936057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5938714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5941354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5943994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5946664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5949342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5951982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5954667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5957346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5960026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5962763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5965485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5968160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5970826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5973497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5976155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5978807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5988153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5990954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5993603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5996252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.5998993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6001720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6004782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6007557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6010178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6012800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6015429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6018087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6020703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6023320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6025981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6028627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6031270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6033903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6036619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6039388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6042072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6044695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6047323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6049958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6052592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6055213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6057842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6060481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6063099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6065718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6068338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6071015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6073756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6076417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6079064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6081698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6084320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6086921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6089548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6092214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6094839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6097490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6100130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6102763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6105646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6108346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6111144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6113857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6116475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6119079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6121697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6124330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6126943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6129566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6132195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6134811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6137423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6140049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6142724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6145455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6148160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6150774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6153418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6156073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6158722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6161361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6164002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6166686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6169320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6171949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6174585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6177269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6179983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6182640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6185299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6187941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6190572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6193226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6195867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6198545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6201164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6203794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6206679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6209328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6212024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6214771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6217466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6220144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6222795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6225430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6228071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6230728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6233373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6236008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6238655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6241293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6243926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6246600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6249322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6252008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6254637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6257274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6259913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6262556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6265189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6267808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6270434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6273093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6275731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6278341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6281015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6283732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6286404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6289025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6291658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6294318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6296942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6299616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6302236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6305086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6307763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6310373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6313000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6315690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6318419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6321088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6323703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6326357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6328984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6331587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6334241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6336901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6339530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6342156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6344794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6347436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6350084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6352765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6355489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6358204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6360864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6363509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6366144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6368805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6371446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6374062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6376692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6379346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6381956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6384572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6387306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6390090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6392768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6395393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6398064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6400712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6403337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6406236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6408875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6411515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6414150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6416766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6419395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6422101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6424868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6427540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6430148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6432781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6435388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6437998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6440638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6443293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6445903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6448522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6451151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6453781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6456438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6459140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6461828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6464451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 62%] 2024-08-06T21:15:49.6467125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6469759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6472406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6475071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6477725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6480372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6483023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6485689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6488337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6491018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6493763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6496449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6499254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6501895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6503318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6504958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6506368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6507785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6509184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6510631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6512025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6513436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6514901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6516438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6517891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6519299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6520729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6522159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6523550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6524967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6526381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6527792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6529202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6530629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6532041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6533483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6534929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6536352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6537808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6539215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6540644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6542035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6543450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6544856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6546248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6547660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6549061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6550503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6551887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6553334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6554820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6556291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6557685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6559098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6560523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6561950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6563344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6564737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6566145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6567544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6568945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6570351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6571800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6573299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6574745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6576135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6577549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6578949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6580371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6581760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6583177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6584583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6585980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6587397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6588811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6590344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6591850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6593324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6594737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6596168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6597604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6599018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6600434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6601855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6603237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6604849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6606349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6607761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6609234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6610761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6612243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6613644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6615048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6616444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6617886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6619307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6620753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6622165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6623604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6625033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6626447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6627922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6629429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6630926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6632333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6633770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6635182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6636618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6638012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6639432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6640856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6642294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6643706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6645115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6646590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6648111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6649568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6650982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6652423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6653852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6655277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6656687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6658130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6659572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6661002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6662420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6663864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6665327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6666776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6668242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6669712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6671152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6672550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6673969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6675384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6676826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6678230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6679667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6681086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6682524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6683925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6685374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6686884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6688347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6689784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6691194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6692633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6694053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6695472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6696880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6698356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6699803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6701222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6702628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6704350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6705905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6707362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6708782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6710219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6711656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6713058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6714476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6715888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6717310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6718705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6720148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6721570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6723061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6724565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6726027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6727466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6728899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6730358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6731774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6733221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6734648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6736077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6737491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6738920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6740363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6741819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6743351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6744834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6746257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6747664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6749094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6750537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6751977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6753386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6754816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6756238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6757683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6759088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6760579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6762084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6763574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6764980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6766411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6767833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6769258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6770702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6772119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6773549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6774965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6776384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6777792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6779266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6780783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6782266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6783671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6785110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6786530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6787928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6789356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6790787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6792218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6793618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6795040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6796454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6797969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6799459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6800942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6802353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6803788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6805414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6806839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6808245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6809669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6811084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6812480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6813911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6815326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6816810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6818332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6819841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6821250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6822661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6824064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6825506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6826930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6828339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6829780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6831209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6832653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6834067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6835534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6837046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6839266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6840691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6842122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6843543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6844977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6846377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6847802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6849224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6850666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6852095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6853509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6854988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6856508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6857972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6859379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6860826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6862248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6863663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6865066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6866503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6867927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6869328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6870775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6872196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6873633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6875084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6876592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6878042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6879474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6880897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6882315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6883726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6885155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6886665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6888097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6889512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6890953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6892365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6893813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6895320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6896777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6898233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6899644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6901103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6902520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6904201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6905670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6907110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6908528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6909937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6911385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6912862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6914407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6915854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6917273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6918689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6920122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6921542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6922964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6924374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6925799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6927197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6928617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6930030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6931516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6933028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6934480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6935914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6937344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6938762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6940171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6941627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6943050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6944466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6945873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6947302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6948715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6950152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6951671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6953128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6954560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6955966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6957388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6958794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6960224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6961639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6963057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6964457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6965901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6967299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6968757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6970255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6971730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6973148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6974554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6975983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6977397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6978813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6980218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6981649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6983052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6984460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6985853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6987318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6988817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6990255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6991682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6993086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6994520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6995913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6997370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.6998802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7000236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7001645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7003065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7004684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7006227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7007752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7009232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7010653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7012077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7013505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7014912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7016339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7017754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7019165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7020583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7022013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7023424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7024839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7026285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7027803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7029264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7030670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7032096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7033512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7034945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7036347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7037770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7039190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7040638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7042041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7043459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7044951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7046472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7047917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7049341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7050766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7052178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7053585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7054985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7056413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7057817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7059227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7060653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7062084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7063541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7065034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7066490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7067923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7069345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7070767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7072186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7073605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7075039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7076441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7077865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7079284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7080740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7082184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7083688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7085133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7086559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7087956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7089368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7090800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7092212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7093631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7095034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7096464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7097918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7099342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7100812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7102323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7103781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7105429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7106837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7108266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7109676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7111097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7112523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7113933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7115362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7116761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7118177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7119636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7121195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7122642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7124045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7125449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7126879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7128267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7129661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7131108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7132516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7133928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7135322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 63%] 2024-08-06T21:15:49.7136752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7138207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7139719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7141187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7142622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7144040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7145465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7146872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7148301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7149727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7151148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7152573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7153980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7155409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7156795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7158251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7159741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7161242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7162641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7164065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7165475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7166907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7168306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7169709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7171159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7172574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7173989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7175391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7176870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7178376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7179843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7181270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7182703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7184118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7185540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7186950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7188372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7189782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7191190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7192605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7194013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7195478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7196960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7198467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7199875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7201324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7202720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7204315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7205788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7207226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7208615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7210009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7211451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7212863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7214336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7215868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7217364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7218780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7220197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7221613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7223027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7224429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7225830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7227219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7228636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7230049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7231440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7232893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7234378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7235845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7237238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7238654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7240069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7241498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7242895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7244316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7245732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7247162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7248565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7249991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7251464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7252965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7254428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7255837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7257261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7258679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7260092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7261490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7262915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7264325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7265739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7267142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7268565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7270034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7271515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7272972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7274373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7275801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7277197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7278611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7280035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7281473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7282866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7284284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7285695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7287124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7288572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7290079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7291541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7292943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7294342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7295735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7297157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7298606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7300040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7301436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7302852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7304504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7305938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7307327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7308797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7310362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7311813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7313229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7314640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7316070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7317462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7318882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7320310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7321736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7323127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7324545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7325940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7327399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7328871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7330369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7331793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7333203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7334612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7336004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7337429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7338837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7340257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7341651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7343077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7344484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7345932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7347410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7348880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7350338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7351745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7353164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7354571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7356003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7357406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7358827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7360249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7361674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7363064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7364511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7365997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7367463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7368855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7370276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7371698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7373103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7374509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7375911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7377329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7378732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7380159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7381550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7383011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7384505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7385957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7387351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7388753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7390204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7391591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7393006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7394405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7395823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7397202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7398654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7400076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7401491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7402919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7405079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7406594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7407991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7409395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7410804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7412223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7413629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7415039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7416431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7417856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7419267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7420695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7422149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7423683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7425141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7426553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7427954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7429351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7430793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7432180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7433581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7434973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7436393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7437783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7439195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7440655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7442156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7443586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7444997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7446399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7447804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7449215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7450636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7452059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7453462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7454873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7456271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7457695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7459145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7460658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7462094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7463509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7464913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7466307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7467708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7469106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7470553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7471935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7473345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7474749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7476169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7477594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7479079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7480549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7481969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7483354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7484770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7486183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7487589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7489001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7490420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7491842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7493256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7494664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7496102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7497657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7499109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7500530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7501911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7503332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7505044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7506485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7507901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7509293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7510735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7512120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7513531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7514924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7516409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7517911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7519376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7520802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7522229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7523622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7525014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7526432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7527841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7529242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7530652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7532063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7533452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7534883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7536345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7537807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7539204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7540636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7542037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7543447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7544879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7546271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7547672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7549072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7550525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7551916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7553386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7554875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7556357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7557750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7559165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7560592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7562019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7563408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7564808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7566221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7567614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7569076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7570487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7571963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7573457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7574906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7576289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7577700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7579105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7580531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7581918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7583317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7584738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7586122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7587534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7588936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7590381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7591817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7593314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7594763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7596194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7597661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7599079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7600498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7601909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7603306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7605020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7606457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7607864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7609268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7610777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7612318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7613784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7615187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7616577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7617999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7619396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7620819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7622209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7623603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7625020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7626407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7627816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7629253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7630829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7632281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7633690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7635078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7636485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7637853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7639249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7640664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7642058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7643453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7644840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7646253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7647690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7649170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7650616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7652032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7653435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7654878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7656270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7657690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7659102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7660515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7661934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7663341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7664765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7666196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7667694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7669134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7670571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7671959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7673360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7674756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7676176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7677566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7678957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7680396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7681802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7683208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7684592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7686045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7687528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7688972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7690376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7691800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7693274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7694673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7696068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7697530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7698930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7700333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7701744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7703130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7704862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7706400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7707856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7709250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7710688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7712069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7713468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7714857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7716272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7717712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7719099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7720536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7721943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7723381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7724918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7726383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7727786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7729190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7730605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7732020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7733421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7734870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7736261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7737651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7739063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7740461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7741909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7743381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7744840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7746215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7747615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7749012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7750447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7751826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7753227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7754627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7756026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7757429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7758814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7760246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7761689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7763180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7764613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7766030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7767497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7768902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7770307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7771717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7773180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7774551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7775948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7777341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7778761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7780209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7781723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7783155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 64%] 2024-08-06T21:15:49.7784572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7785957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7787359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7788744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7790165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7791550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7792944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7794338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7795733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7797133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7798645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7800184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7801653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7803061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7804755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7806217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7807609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7808997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7810378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7811787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7813177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7814551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7815945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7817325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7818819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7820336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7821810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7823203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7824616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7825996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7827396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7828793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7830216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7831600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7832990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7834401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7835797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7837267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7838810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7840333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7841729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7843125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7844503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7845908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7847308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7848707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7850085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7851478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7852898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7854276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7855730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7857230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7858708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7860101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7861523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7862936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7864367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7865766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7867189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7868607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7870065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7871468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7872873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7874336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7875829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7877289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7878684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7880122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7881541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7882950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7884351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7885777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7887185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7888599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7890012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7891423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7892893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7894369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7895837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7897249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7898730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7900635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7902116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7903529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7905285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7906717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7908141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7909551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7910999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7912514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7914046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7915550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7916962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7918383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7919801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7921246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7922664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7924080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7925482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7927026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7928445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7929857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7931271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7932723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7934297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7935737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7937156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7938567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7939995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7941412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7942834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7944233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7945655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7947041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7948454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7949852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7951325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7953086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7955083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7956837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7958278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7959730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7961130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7962591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7964008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7965421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7966823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7968260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7969676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7971165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7972689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7974155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7975596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7977004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7978435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7979842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7981268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7982681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7984092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7985500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7986930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7988326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7989780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7991321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7992820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7994218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7995642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7997083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7998544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.7999961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8001364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8002818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8004529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8005989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8007395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8008912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8010453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8011931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8013357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8014765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8016190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8017577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8018995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8020407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8021839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8023258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8024679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8026092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8027562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8029041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8030511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8031922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8033425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8034828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8036246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8037680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8039116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8040514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8041935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8043369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8044839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8046280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8047793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8049241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8050640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8052043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8053454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8054877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8056289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8057771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8060579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8062064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8063483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8064907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8066297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8068330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8071557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8073040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8074463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8075887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8077999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8080461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8083737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8085205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8086644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8088493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8091423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8092884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8096400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8100515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8103385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8106193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8107653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8110177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8111891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8114575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8116172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8117754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8119763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8121444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8123864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8125429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8127182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8129508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8131332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8132902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8134327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8135754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8137190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8138587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8140003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8141398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8142821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8144204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8145671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8147161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8148632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8150018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8151411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8152830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8154231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8155642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8157038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8158456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8159868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8161281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8162677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8164164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8165671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8167135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8168532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8169963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8171373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8172762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8174184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8175602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8177028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8178407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8179816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8181221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8182694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8184172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8185648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8187050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8188477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8189869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8191265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8192684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8194096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8195516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8196916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8198392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8199807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8201264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8202746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8204542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8205996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8207409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8208814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8210230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8211634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8213027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8214434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8215856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8217286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8218679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8220083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8221571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8223112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8224584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8226016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8227421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8228956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8230343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8231736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8233161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8234568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8235999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8237398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8238821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8240267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8241757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8243196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8244605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8246026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8247428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8248808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8250211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8251629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8253010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8254416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8255833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8257256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8258718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8260205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8261652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8263080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8264474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8265908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8267311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8268767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8270165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8271568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8272992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8274405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8276339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8277836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8279336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8280782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8282186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8283576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8285031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8286480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8287936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8289362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8290772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8292203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8293598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8295012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8296484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8298061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8299508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8300899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8302327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8303741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8305472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8306903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8308333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8309747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8311158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8312560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8313977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8315448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8316991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8318443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8319839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8321266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8322662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8324078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8325480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8326934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8328389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8329871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8331283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8332710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8334099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8335556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8337076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8338555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8339952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8341359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8342789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8344204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8345620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8347048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8348466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8349868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8351270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8352659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8354115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8355606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8357080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8358475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8359942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8361366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8362756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8364174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8365580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8367062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8368456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8369867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8371271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8372742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8374236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8375702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8377132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8378624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8380014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8381408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8382818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8384213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8385612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8387270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8389391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8391464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8393352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8394861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8396960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8399232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8401404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8403482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8406038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8408117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8409990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8411916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8414002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8415779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8417342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8419273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8421648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8423559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8425708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8428100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8429969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8431906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8433755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8435567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8437324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8439241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8441149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8443060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8444892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8446562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8448440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8450088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8451728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8453310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8455090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8457147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8459035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8460661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8462247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8463985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8465599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8467245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8468754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8470268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8471716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8473145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8474536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8475950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8477375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8478788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8480203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 65%] 2024-08-06T21:15:49.8481597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8483019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8484432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8485843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8487304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8488809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8490259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8491662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8493056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8494479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8495887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8497315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8498772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8500173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8501599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8502996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8504921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8506459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8508031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8509481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8510885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8512287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8513710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8515093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8516505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8517927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8519330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8520736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8522132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8523564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8525020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8526526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8528009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8529438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8530852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8532269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8533675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8535101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8536518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8537932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8539356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8540755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8542175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8543602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8545095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8546546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8547989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8549458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8550876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8552283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8553713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8555110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8556527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8557948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8559358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8560765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8562155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8563620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8565114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8566618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8568033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8569457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8570867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8572275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8573675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8575086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8576483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8577879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8579282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8580678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8582140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8583610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8585061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8586461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8587898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8589283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8590689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8592095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8593518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8594905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8596294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8597780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8599208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8600670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8602170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8603648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8605355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8606789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8608202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8609612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8611009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8612417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8613809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8615239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8617108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8618565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8620064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8621591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8623077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8624459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8625891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8627300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8628738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8630124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8631537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8632940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8634366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8635761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8637157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8638633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8640123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8641579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8642976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8644392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8645795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8647198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8648605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8650026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8651430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8652832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8654228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8655621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8657035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8658477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8659970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8661409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8662830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8664213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8665618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8667015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8668456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8669846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8671247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8672649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8674061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8675465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8676895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8678417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8679867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8681262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8682652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8684061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8685454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8686854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8688255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8689663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8691063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8692463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8693846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8695289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8696794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0004s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8698297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8699716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8701125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8702554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8704221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8705717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8707143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8708653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8710047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8711465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8712865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8714340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8715861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8717309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8718742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8720149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8721562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8722951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8724365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8725776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8727178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8728588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8730014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8731424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8732809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8734261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8735754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8737235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8738661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8740083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8741490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8742919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8744315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8745732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8747127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8748565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8749951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8751353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8752808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8754300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8755748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8757144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8758587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8759992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8761402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8762794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8764216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8765614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8767018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8768433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8769859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8771311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8772788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8774247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8775656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8783949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8785603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8787022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8788431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8789862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8791248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8792658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8794065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8795508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8797001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8798605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8800886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8802318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8803710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8805446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8806875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8808283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8809691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8811080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8812502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8813911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8815317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8816827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8818377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8819932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8821332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8822726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8824142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8825540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8826938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8828346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8829737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8831155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8832539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8833946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8835333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8836813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8838280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8839734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8841135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8842568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8843963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8845359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8846805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8848218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8849627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8851027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8852452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8853866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8855322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8856830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8858291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8859694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8861100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8862487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8863902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8865309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8866721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8868126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8869521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8870943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8872331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8873786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8875316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8876814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8878280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8879694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8881103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8882530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8883925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8885326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8886769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8888183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8889601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8891003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8892461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8893962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8895415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8896821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8898280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8899701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8901113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8903003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8904881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8906344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8907753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8909165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8910566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8912088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8913597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8915067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8916468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8917914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8919307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8920719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8922118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8923552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8924941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8926334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8927765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8929167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8930557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8931983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8933481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8934926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8936330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8937739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8939151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8940556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8941956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8943341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8944744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8946167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8947586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8949002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8950451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8951967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8953406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8954825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8956232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8957853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8959505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8960938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8962342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8963741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8965144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8966533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8967971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8969445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8970958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8972388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8973805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8975210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8976611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8978014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8979428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8980831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8982210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8983610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8985016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8986434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8987879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8989370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8990823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8992237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8993619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8995029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8996510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8997987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.8999371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9000771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9002162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9003564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9005468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9006934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9008477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9010009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9011468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9012855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9014274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9015672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9017069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9018485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9019910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9021311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9022701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9024113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9025511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9026972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9028466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9029918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9031308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9032722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9034100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9035490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9036879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9038321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9039702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9041084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9042501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9043903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9045356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9046830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9048321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9049726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9051203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9052589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9054002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9055405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9056810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9058223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9059623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9061047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9062431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9063901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9065406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9066874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9068269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9069671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9071070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9072486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9073865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9075269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9076662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9078083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9079486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9080869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9082278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9083720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9085205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9086634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9088075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9089490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9090894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9092282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9093705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9095112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9096508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9097953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9099356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9100762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9102176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9103657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9105755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9107192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9108593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9109997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9111382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9112791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9114173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9115568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9116964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9118390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9119795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9121269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9122808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9124275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9125677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9127068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9128508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9129912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9131320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9132736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9134146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9135536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9136915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9138338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 66%] 2024-08-06T21:15:49.9139777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9141278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9142704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9144102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9145494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9146912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9148310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9149712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9151152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9152613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9154031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9155458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9156912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9158372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9159857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9161381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9162886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9164341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9165787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9167217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9168703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9170142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9171576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9172992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9174443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9175881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9177298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9178800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9180374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9181869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9183290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9184733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9186163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9187616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9189061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9190496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9191928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9193389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9194812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9196249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9197719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9199312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9200802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9202228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9203677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9205555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9207007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9208427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9209889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9211319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9212753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9214170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9215611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9217138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9218712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9220194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9221618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9223070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9224486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9225915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9227349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9228808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9230221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9231660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9233086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9234532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9235986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9237503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9238983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9240406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9241832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9243241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9244677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9246103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9247532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9248966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9250454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9251880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9253304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9254800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9256358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9257834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9259299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9260727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9262163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9263619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9265072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9266523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9267975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9269455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9270900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9272341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9273827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9275399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9276850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9278279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9279731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9281183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9282596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9284044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9285496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9286927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9288389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9289827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9291293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9292775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9294299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9295765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9297219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9298710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9300171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9301599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9303054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9304969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9306523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9308005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9309459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9310903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9312414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9313976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9315466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9316919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9318337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9319788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9321222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9322677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9324096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9325513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9326958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9328386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9329842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9331315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9332848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9334335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9335770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9337189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9338638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9340089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9341524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9342942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9344374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9345798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9347202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9348624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9350113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9351646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9353101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9354616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9356620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9358101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9359507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9360963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9362398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9363849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9365271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9366713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9368141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9369653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9371212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9372682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9374133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9375575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9377018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9378442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9379889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9381332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9382758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9384176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9385666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9387095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9388572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9390079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9391584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9393035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9394455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9395891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9397312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9398808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9400234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9401669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9403100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9404993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9406433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9407978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9409538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9411048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9412490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9413915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9415355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9416781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9418209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9419621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9421076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9422506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9423935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9425342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9426825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9428339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9429845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9431280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9432719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9434171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9435585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9437021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9438454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9439902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9441337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9442781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9444212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9445702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9447204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9448686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9450108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9451572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9452978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9454393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9455834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9457258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9458685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9460525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9462021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9463451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9464948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9466466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9467957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9469385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9470844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9472282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9473793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9475250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9476664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9478094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9479517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9480969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9482401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9483917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9485419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9486906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9488315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9489740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9491159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9492614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9494029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9495446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9496876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9498346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9499782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9501194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9502678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9504671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9506207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9507622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9509073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9510505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9511939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9513361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9514805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9516268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9517683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9519123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9520546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9522061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9523567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9525041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9526465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9527907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9529318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9530744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9532180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9533630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9535041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9536456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9537904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9539337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9540811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9542351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9543847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9545285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9546727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9548150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9549592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9551030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9552490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9553911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9555354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9556785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9558193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9559685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9561192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9562701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9564119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9565559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9566982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9568426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9569833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9571265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9572710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9574155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9575568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9577004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9578473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9579992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9581464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9582905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9584358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9585796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9587222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9588638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9590078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9591498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9592934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9594344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9595777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9597242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9598797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9600276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9601698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9603157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9604996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9606495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9607923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9609369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9610779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9612210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9613642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9615093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9616621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9618183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9619668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9621099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9622538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9623964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9625393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9626809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9628239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9629647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9631083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9632518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9633946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9635396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9636915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9638385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9639810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9641225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9642659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9644094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9645496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9646922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9648341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9649775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9651185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9652633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9654091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9655611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9657060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9658484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9659896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9661313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9662755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9664156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9665594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9667016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9668430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9669836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9671267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9672750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9674255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9675717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9677157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9678583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9680016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9681436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9682885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9684340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9685754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9687193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9688622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9690070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9691523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9693066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9694528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9695963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9697365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9698840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9700335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9701775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9703208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9705070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9706541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9707968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9709396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9710898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9712460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9713966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9715398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9716812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9718259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9719691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9721123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9722548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9723995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9725452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9726878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9728332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9729825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9731367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9732844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9734276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9735699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9737223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9738642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9740086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9741518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9742983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9744410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9745837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9747288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9748741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9750241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9751769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9753292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9754739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9756179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9757595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9759035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9760468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9761907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9763353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9764804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9766241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9767645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9769157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9771508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9773050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9774483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9775996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9777413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9778848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9780256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9781682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9783104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9784576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9785991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9787405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9788899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9790420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9791900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9793323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9794796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9796242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9797676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9799149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9800591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 67%] 2024-08-06T21:15:49.9802011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9803439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9805378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9806850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9808396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9809941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9811429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9812846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9814295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9815727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9817157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9818577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9820014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9821416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9822843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9824265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9825724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9827168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9828697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9830171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9831600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9833030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9834442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9835882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9837299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9838714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9840109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9841541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9842962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9844380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9845829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9847345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9848808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9850208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9851631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9853046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9854516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9855930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9857354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9858769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9860204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9861609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9863042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9864511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9866036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9867487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9868914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9870328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9871737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9873146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9874562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9875993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9877406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9878829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9880234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9881662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9883113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9884634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9886089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9887522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9888944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9890354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9891775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9893194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9894660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9896073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9897502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9898971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9900414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9901876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9903638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9905767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9907228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9908631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9910057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9911470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9912894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9914323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9915755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9917182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9918603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9920085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9921623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9923185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9924666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9926104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9927535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9928971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9930392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9931798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9933230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9934650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9936117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9937537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9938973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9940442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9941968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9943411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9944829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9946264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9947699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9949165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9950587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9952005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9953425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9954841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9956269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9958292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9959920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9961888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9963420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9964872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9966299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9967747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9969157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9970592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9972014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9973435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9974840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9976269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9977692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9979146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9980628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9982107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9983521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9984919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9986345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9987773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9989203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9990608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9992031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9993462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9994917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9996339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9997777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:49.9999315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0000865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0002329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0003773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0005722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0007180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0008621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0010049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0011491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0012918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0014353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0015768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0017241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0018801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0020373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0021855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0023303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0024737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0026169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0027614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0029042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0030495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0031918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0033359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0034792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0036245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0037729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0039266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0040749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0042206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0043631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0045074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0046497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0047951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0049382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0050801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0052247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0053717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0055171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0056682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0058217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0059719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0061147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0062613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0064049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0065500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0066912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0068361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0069790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0071226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0072660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0074071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0075554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0077118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0078609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0080031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0081474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0082898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0084320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0085738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0087189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0088615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0090046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0091459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0092875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0094357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0095853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0097335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0098819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0100285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0101706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0103143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0105054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0106543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0107981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0109428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0110861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0112312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0113824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0115395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0116896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0118353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0119789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0121209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0122656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0124092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0125528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0126949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0128422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0129858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0131287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0132751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0134290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0135777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0137201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0138670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0140107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0141565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0142993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0144437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0145870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0147327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0148771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0150209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0151679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0153206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0154660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0156093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0157531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0159134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0160920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0163712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0166408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0169096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0171797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0174486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0177222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0179996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0182785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0185483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0188189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0190896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0193633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0196317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0199087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0201801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0204942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0207707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0210393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0213087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0216187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0218992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0221711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0224402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0227085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0229753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0232454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0235165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0237828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0240546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0243259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0245971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0248665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0251428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0254223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0256973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0259677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0262395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0265091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0267793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0270516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0273189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0275877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0278574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0281250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0283914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0286636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0289519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0292247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0294930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0297627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0300382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0303067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0307748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0312512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0317868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0323559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0329190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0334917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0340825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0346683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0352461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0358122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0363744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0369429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0375043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0380681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0386320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0391987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0397576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0403344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0409339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0415157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0421005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0426751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0432395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0438029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0443625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0449266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0454927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0460567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0466144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0471794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0477448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0483118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0488864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0494733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0500500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0506566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0512241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0517881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0523542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0529175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0534714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0540355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0546036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0551699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0557276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0563177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0569092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0574765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0580361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0585854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0590597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0595451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0600813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0606289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0611540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0614981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0617701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0620421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0623301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0626155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0628895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0631604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0634290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0636958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0639620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0642288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0644974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0647641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0650318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0652969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0655673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0658391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0661150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0663859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0666537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0669210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0671852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0674556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0677272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0679953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0682615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0685295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0687977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0690659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0693379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0696161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0698942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0701615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0704800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0707509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0710193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0712878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0715540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0718208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0720876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0723565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0726226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0729005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0731909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0734654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0737311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0739997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0742694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0745376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0748062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0750772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0753479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0756146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0758851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0761538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0764275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0767067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0769786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0772447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0775152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0777853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0780515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0783201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0785974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0788644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0791309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0793992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0796684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0799476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0802239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0805482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0808262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0810951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0813630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0816311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0818999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0821704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0824498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0827176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0829863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0832546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0835317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0838103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0840854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0843543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0846209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0848881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0851581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0854228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0856892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0859650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0862357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0865039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0867714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0870418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0873211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0875990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0878706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0881417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0884112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0886773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 68%] 2024-08-06T21:15:50.0889440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0892131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0894811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0897475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0900225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0902905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0906506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0909333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0912055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0914734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0917418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0920098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0922737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0925412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0928105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0930758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0933433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0936127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0938798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0941521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0944265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0946986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0949673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0952363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0955016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0957678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0960367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0963025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0965668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0968336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0971011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0973661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0976359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0979120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0981837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0984491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0987160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0989818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0992525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0995347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.0998037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1000776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1003477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1006654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1009327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1012136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1014982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1017737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1020406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1023058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1025741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1028403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1031083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1033787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1036479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1039150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1041798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1044465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1047185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1049968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1052668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1055342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1058011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1060720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1063391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1066066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1068790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1071476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1074126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1076816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1079517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1082248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1085014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1087777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1090495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1093153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1095811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1098483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1101208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1103879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1107080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1109741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1112428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1115119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1117772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1120585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1123393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1126107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1128805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1131481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1134191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1136864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1139526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1142195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1144880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1147552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1150202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1152863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1155568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1158297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1161018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1163676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1166351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1169022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1171676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1174323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1176996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1179671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1182328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1185006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1187713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1190437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1193175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1195909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1198668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1201369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1204459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1207231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1209919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1212601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1215296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1218003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1220693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1223352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1226075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1228864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1231607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1234272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1236980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1239647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1242337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1245007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1247670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1250338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1253023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1256102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1258790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1261524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1264301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1267014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1269696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1272368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1275070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1277719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1280384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1283046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1285708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1288362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1290999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1293660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1296727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1299577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1302278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1305361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1308093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1310733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1313396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1316083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1318760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1321539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1324206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1326886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1329554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1332333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1335109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1338378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1341113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1343812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1346446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1349101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1351782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1354431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1357069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1359724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1362405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1365062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1367778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1370546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1374177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1376834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1379468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1382126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1384796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1387478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1390136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1392812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1395469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1398162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1400863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1403565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1406902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1409652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1412312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1414965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1417620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1420307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1422947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1425641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1428316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1430956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1433620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1436286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1439036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1441813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1444528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1447179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1449861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1452542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1455190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1457862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1460555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1463224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1465891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1468566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1471246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1473956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1476706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1479420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1482081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1484737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1487376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1490031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1492701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1495362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1497990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1500715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1503385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1506532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1509264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1512040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1514762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1517425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1520098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1522786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1525475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1528142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1530785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1533482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1536160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1538854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1541516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1544219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1546965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1549676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1552310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1554965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1557666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1560350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1563000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1565650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1568329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1570996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1573653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1576341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1579013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1581705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1584449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1587188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1589876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1592546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1595223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1597872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1600603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1603271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1606384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1609067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1621227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1624143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1627001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1629832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1632571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1635255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1637901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1640580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1643240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1645896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1648536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1651193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1653866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1656528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1659167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1661872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1664643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1667369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1670021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1672691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1675376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1678016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1680683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1683342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1685995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1688648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1691284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1692693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1694170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1695680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1697135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1698560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1700048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1701462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1702858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1704786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1706291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1707705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1709100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1710534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1711945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1713460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1715002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1716478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1717910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1719384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1720799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1722194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1723615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1725014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1726493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1727890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1729320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1730710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1732162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1733648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1735115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1736524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1737917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1739348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1740764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1742179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1743586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1745034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1746458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1747883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1749292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1750769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1752277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1753752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1755175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1756587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1758023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1759416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1760832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1762245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1763674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1765088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1766509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1767918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1769442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1770943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1772456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1773864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1775299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1776728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1778134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1779571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1780993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1782416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1783825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1785285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1786709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1788123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1789573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1791085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1792545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1793955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1795382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1796792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1798226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1799676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1801110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1802521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1804333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1805832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1807252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1808742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1810302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1811756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1813173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1814587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1816022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1817435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1818837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1820256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1821659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1823072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1824473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1825941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1827382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1828868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1830300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1831722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1833130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1834534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1835948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1837350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1838774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1840155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1841565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1842979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1844412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1845885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1847392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1848856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1850291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1851698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1853126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1854546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1855984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1857404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1858813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1860234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1861644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1863052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1864492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1866025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1867483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1868897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1870300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1871727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1873140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1874553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1875975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1877380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1878808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1880205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1881618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1883064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1884579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1886039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1887452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1888928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1890364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1891760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1893179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1894582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1896005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1897407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1898840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1900267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1901710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1903204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1905152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1907023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1908467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1909881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1911275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1912696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1914108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1915499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1916942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1918353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1919781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1921168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1922699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 69%] 2024-08-06T21:15:50.1924246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1925739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1927157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1928582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1929981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1931401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1932786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1934211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1935614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1937036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1938443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1939829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1941291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1942787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1944246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1945642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1947079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1948487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1949901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1951298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1952721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1954129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1955525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1956960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1958367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1959835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1961314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1962776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1964174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1965598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1966990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1968388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1969851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1971283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1972673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1974084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1975488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1976897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1978346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1979845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1981312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1982719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1984126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1985515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1986934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1988342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1989752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1991150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1992569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1993985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1995386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1996844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1998328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.1999846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2001239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2002643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2004468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2005971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2007362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2008801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2010198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2011615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2013006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2014434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2015851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2017370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2018906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2020361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2021787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2023201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2024613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2026029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2027461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2028871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2030283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2031693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2033117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2034534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2036094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2037614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2039080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2040526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2041961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2043377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2044785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2046293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2047702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2049124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2050553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2052022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2053443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2054956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2056559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2058079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2059541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2060964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2062413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2063844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2065283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2066728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2068162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2069607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2071070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2072495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2074003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2075567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2077060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2078487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2079909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2081350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2082757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2084190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2085611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2087072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2088477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2089916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2091339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2092832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2094332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2095804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2097268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2098704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2100184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2101600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2103039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2105010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2106467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2107898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2109337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2110759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2112308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2113858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2115360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2116783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2118212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2119641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2121056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2122515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2123916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2125338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2126748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2128211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2129616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2131096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2132599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2134086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2135495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2136926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2138351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2139758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2141170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2142569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2144004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2145420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2146841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2148241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2149751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2151249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2152710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2154113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2155554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2156997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2158403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2159832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2161259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2162706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2164116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2165554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2166986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2168479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2169973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2171446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2172859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2174297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2175697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2177132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2178553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2179977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2181396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2182816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2184256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2185680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2187169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2188663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2190162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2191597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2193019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2194432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2195868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2197320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2198753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2200209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2201634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2203085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2204913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2206465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2208018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2209517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2210916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2212340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2213755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2215188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2216592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2218044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2219460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2220878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2222303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2223712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2225181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2226695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2228191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2229593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2231025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2232446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2233864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2235267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2236702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2238141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2239563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2240976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2242382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2243848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2245329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2246801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2248233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2249673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2251079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2252499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2253906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2255338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2256734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2258166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2259589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2261035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2262481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2263997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2265490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2266942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2268415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2269834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2271274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2272704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2274138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2275552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2276989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2278409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2279826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2281274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2282772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2284255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2285657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2287099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2288518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2289956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2291356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2292781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2294192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2295637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2297066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2298486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2300047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2301605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2303104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2304980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2306485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2307936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2309361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2310769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2312197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2313609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2315019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2316421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2317873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2319283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2320954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2322525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2324009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2325441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2326846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2328287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2329700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2331136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2332534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2333957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2335405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2336867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2338305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2339819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2341364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2342858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2344264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2345675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2347095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2348524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2349940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2351340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2352768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2354183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2355599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2356996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2358489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2359991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2361452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2362859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2364293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2365727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2367172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2368694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2370148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2371605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2373038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2374479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2375906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2377428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2379062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2380558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2381975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2383412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2384812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2386215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2387644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2389064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2390483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2391889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2393327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2394756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2396274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2397800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2399368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2400843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2402289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2403744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2405811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2407269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2408748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2410313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2411851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2413387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2414817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2416610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2418230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2419749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2421140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2422550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2423954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2425450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2426843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2428245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2429688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2431107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2432523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2433922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2435422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2436973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2438466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2439898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2441335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2442756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2444174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2445580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2447018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2448438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2449868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2451294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2452696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2454166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2455646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2457113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2458522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2459980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2461381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2462801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2464210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2465646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2467041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2468442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2469905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2471329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2472821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2474378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2475947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2477376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2478806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2480244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2481680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2483102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2484533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2485948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2487372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2488786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2490198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2491670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2493975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2495452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2496849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2498272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2499804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2501264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2502670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2504650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2506141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2507585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2509037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2510515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2512361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2514160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2515773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2517267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2518731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2520159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2521621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2523071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2524505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2525915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2527325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2528728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2530159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2531648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2533184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2534664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2536077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2537509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2538923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2540363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2541792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2543239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2544638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2546056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2547478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2548907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2550302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2551810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2553359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2554831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2556251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2557657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2559080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2560486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2561904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2563301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2564733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2566167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2567606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2569022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2570513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2572047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2573509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2574922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2576344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2577783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2579189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2580613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2582049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2583492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 70%] 2024-08-06T21:15:50.2584892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2586326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2587747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2589233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2590730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2592224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2593681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2595111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2596553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2598009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2599540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2601055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2602569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2604434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2606195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2607721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2609593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2611174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2612699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2614161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2615584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2617011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2618429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2619874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2621280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2622706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2624145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2625590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2626998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2628478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2630024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2631545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2632952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2634417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2635836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2637258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2638681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2640090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2641518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2642931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2644371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2645782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2647264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2648778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2650249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2651645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2653077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2654515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2655924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2657339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2658762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2660201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2661602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2663101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2664522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2665996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2667467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2668929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2670337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2671768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2673157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2674586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2675989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2677399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2678810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2680203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2681639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2683058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2684628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2686126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2687602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2689019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2690442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2691853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2693288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2694723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2696151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2697558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2698963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2700451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2701848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2703330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2705362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2706918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2708321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2709748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2711163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2712596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2714000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2715426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2716834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2718249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2719667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2721068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2722564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2724101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2725574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2726975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2728402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2729823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2731242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2732643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2734076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2735485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2736889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2738286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2739691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2741117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2742550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2744070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2745529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2746955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2748347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2749758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2751169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2752601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2754022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2755441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2756864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2758286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2759698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2761143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2762659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2764161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2765587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2766993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2768416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2769821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2771231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2772623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2774064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2775472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2776883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2778281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2779726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2781238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2782676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2784116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2785576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2787003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2788397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2789822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2791234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2792656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2794075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2795499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2796912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2798371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2799915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2801378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2802800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2804663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2806118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2807511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2808936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2810344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2811751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2813142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2814591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2816001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2817495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2819010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2820472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2821899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2823297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2824734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2826141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2827571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2828967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2830383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2831787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2833220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2834634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2836093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2837599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2839049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2840449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2841837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2843251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2844673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2846081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2847474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2848895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2850301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2851717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2853110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2854607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2856104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2857546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2858958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2860369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2861808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2863206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2864639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2866051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2867479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2868873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2870296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2871694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2873153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2874646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2876106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2877506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2878913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2880326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2881725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2883146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2884572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2885982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2887375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2888809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2890226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2891634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2893077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2894610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2896079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2897483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2898912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2900369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2901810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2903217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2905062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2906551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2908369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2909799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2911209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2912709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2914293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2915755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2917177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2918591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2920002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2921417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2922820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2924267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2925679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2927097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2928499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2929931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2931395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2932904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2934365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2935805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2937232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2938635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2940060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2941464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2942887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2944282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2945699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2947113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2948541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2949977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2951489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2952942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2954386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2955787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2957201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2958599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2960005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2961419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2962809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2964249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2965665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2967078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2968520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2970085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2971569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2972985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2974411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2975829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2977226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2978609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2980020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2981418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2982845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2984263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2985672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2987129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2988644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2990077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2991483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2992891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2994349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2995751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2997171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.2998591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3000057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3001482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3002892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3004709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3006238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3007783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3009249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3010665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3012071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3013484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3014904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3016328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3017743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3019142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3020561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3021976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3023409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3024823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3026283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3027779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3029256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3030655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3032080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3033494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3034949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3036354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3037776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3039188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3040612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3042033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3043432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3044923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3046420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3047892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3049289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3050716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3052130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3053541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3054965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3056390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3057803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3059193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3060616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3062022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3063492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3064984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3066444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3067850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3069284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3070680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3072097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3073505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3074954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3076344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3077760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3079163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3080564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3082009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3083475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3084954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3086388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3087803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3089196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3090608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3092010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3093421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3094832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3096263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3097677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3099073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3100573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3102073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3103551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3105342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3106796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3108211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3109640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3111041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3112472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3113878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3115322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3116718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3118115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3119637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3121190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3122666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3124064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3125503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3126923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3128337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3129739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3131167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3132581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3134010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3135426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3136856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3138312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3139794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3141254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3142664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3144094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3145504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3146945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3148344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3149764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3151146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3152546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3153946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3155373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3156756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3158185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3159683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3161138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3162544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3163939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3165363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3166773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3168182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3169582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3171005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3172417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3173837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3175235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3176697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3178191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3179627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3181043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3182441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3183869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3185259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3186668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3188067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3189487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3190874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3192287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3193691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3195157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3196643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3198084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3199539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3200952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3202362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3203761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3205565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3207025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3208434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3209830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3211258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3212670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3214161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3215694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3217149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3218635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3220019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3221429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3222828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3224263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3225660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3227070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3228471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3229897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3231288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3232737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3234250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3235724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3237115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 71%] 2024-08-06T21:15:50.3238514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3239938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3241340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3242747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3244179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3245601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3247008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3248422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3249819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3251271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3252761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3254232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3255619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3257013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3258438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3259827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3261234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3262659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3264107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3265492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3266897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3268305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3269726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3271168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3272693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3274176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3275609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3277012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3278411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3279835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3281252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3282668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3284088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3285507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3286907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3288306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3289734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3291240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3292685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3294110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3295505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3296904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3298328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3299760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3301175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3302596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3304435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3305885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3307316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3308810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3310380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3311853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3313294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3314746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3316193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3317606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3319025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3320458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3321873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3323292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3324718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3326155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3327617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3329131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3330601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3332040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3333466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3334889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3336298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3337721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3339165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3340570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3341997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3343430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3344883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3346337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3347860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3349336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3350786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3352208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3353654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3355069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3356515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3357918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3359325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3360764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3362191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3363633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3365085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3366602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3368070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3369492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3370901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3372333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3373772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3375196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3376610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3378050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3379473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3380883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3382318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3383806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3385333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3386786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3388217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3389636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3391073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3392466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3393912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3395336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3396772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3398176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3399623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3401061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3402520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3404421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3405940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3407379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3408810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3410241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3411648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3413087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3414540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3415969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3417382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3418831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3420261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3421725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3423263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3424746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3426186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3427591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3429016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3430428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3431871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3433281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3434721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3436135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3437575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3438980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3440432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3441958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3444148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3445586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3447003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3448445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3449870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3451297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3452716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3454182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3455612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3457043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3458453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3459927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3461427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3462874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3464302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3465721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3467161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3468563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3469988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3471409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3472848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3474273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3475701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3477118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3478597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3480100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3481573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3482990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3484431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3485863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3487275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3488707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3490132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3491552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3492963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3494412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3495830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3497278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3498759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3500280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3501699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3503102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3504916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3506402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3507841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3509246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3510673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3512094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3513543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3514979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3516503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3518062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3519583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3521000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3522434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3523864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3525316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3526751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3528174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3529610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3531029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3532453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3533863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3535364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3536885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3538354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3539761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3541198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3542626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3544052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3545478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3546901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3548340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3549743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3551174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3552597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3554034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3555509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3557033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3558500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3559944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3561354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3562784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3564190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3565611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3567023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3568422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3569860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3571273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3572688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3574138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3575696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3577157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3578577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3579979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3581413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3582832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3584265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3585679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3587097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3588537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3589940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3591364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3592822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3594363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3595822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3597251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3598661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3600132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3601537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3602963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3604763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3606214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3607634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3609040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3610463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3611951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3613519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3615024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3616460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3617881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3619302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3620708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3622144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3623568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3625017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3626424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3627848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3629290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3630734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3632249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3633710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3635165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3636575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3637995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3639404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3640842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3642243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3643661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3645093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3646534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3647934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3649375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3650889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3652356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3653771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3655200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3656638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3658056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3659480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3660888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3662314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3663729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3665179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3666588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3668029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3669537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3670977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3672396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3673809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3675269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3676661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3678080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3679487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3680917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3682311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3683727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3685156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3686629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3688129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3689586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3691029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3692454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3693873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3695290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3696726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3698152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3699611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3701020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3702443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3703853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3705738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3707258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3708731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3710163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3711564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3712988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3714413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3715850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3717256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3718680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3720092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3721532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3722936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3724404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3725907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3727391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3728788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3730195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3731630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3733050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3734486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3735903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3737328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3738740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3740155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3741560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3743026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3744551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3746018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3747420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3748830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3750263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3751662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3753085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3754522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3755963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3757370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3758795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3760216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3761694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3763199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3764688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3766104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3767544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3768953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3770365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3771792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3773207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3774645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3776049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3777488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3778903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3780311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3781755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3783280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3784768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3786182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3787590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3789012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3790421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3791820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3793237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3794689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3796125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3797523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3798939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3800426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3801952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3803392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3805196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3806669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3808097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3809482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3810879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3812302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3813703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3815132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3816528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3817949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3819427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3820958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3822413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3823850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3825286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3826700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3828102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3829543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3830959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3832359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3833783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3835203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3836643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3838099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3839617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3841070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3842497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3843900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3845316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3846723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3848160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3849559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3850955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3852380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3853808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3855226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3856669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3858184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3859645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3861063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3862469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3863925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3865348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3866769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3868174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3869604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3871024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3872429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3873880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3875365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3876886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3878325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3879742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3881159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3882588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3884008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3885427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3886840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3888273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3889669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3891074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3892498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 72%] 2024-08-06T21:15:50.3893970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3895474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3896939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3898368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3899824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3901245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3902647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3904455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3905907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3907326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3908729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3910496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3911926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3913403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3914956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3916422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3917846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3919239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3920660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3922066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3923494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3924909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3926324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3927734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3929173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3930572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3932019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3933541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3935031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3936455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3937861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3939301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3940723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3942144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3943555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3944992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3946399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3947806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3949197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3950664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3952167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3953608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3955041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3956450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3957885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3959283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3960704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3962112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3963540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3964945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3966364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3967774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3969200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3970640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3972142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3973618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3975041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3976465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3977870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3979290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3980697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3982103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3983495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3984923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3986332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3987740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3989172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3990679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3992129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3993546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3994966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3996377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3997802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.3999201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4000660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4002080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4003530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4005304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4006746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4008229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4009775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4011236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4012659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4014074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4015487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4016891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4018286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4019706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4021110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4022514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4023929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4025350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4026798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4028290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4029729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4031157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4032571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4033989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4035410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4036827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4038264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4039658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4041087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4042500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4043958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4045393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4046925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4048371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4049793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4051182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4052574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4054014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4055424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4056841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4058239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4059659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4061074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4062483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4063941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4065445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4066897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4068304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4069700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4071124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4072527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4073945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4075370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4076775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4078203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4079601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4081017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4082453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4083984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4085417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4086814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4088209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4089634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4091018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4092407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4093853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4095259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4096666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4098065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4099525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4100980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4102475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4104241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4105746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4107203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4108655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4110059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4111484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4112899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4114320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4115743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4117144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4118565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4119945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4121435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4122958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4124468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4125863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4127279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4128681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4130108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4131505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4132905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4134369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4135801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4137240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4138655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4140144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4141665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4143145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4144583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4146032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4147462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4148891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4150359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4151794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4153214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4154635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4156064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4157479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4158962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4160456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4161979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4163401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4164855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4166263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4167691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4169116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4170560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4171975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4173414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4174838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4176265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4177739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4179239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4180738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4182173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4183618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4185037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4186471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4187897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4189318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4190725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4192162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4193609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4195017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4196488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4197990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4199514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4200931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4202366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4203805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4205618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4207025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4208449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4209861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4211298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4212703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4214151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4215643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4217181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4218668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4220080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4221504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4222916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4224348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4225749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4227174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4228583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4229998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4231404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4232832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4234331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4235843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4237297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4238719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4240165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4241575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4243003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4244453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4245898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4247309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4248740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4250163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4251611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4253066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4254612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4256071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4257491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4258913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4260322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4261757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4263183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4264622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4266025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4267460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4268885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4270306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4271753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4273278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4274767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4276198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4277616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4279045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4280491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4281903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4283373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4284799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4286249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4287670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4289104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4290558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4292082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4293534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4294965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4296387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4297828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4299228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4300674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4302115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4303559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4305358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4306796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4308233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4309723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4311266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4312730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4314192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4315620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4317047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4318459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4319878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4321321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4322730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4324180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4325598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4327028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4328462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4329987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4331446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4332884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4334313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4335738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4337144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4338579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4340712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4342156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4343596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4345047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4346471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4347946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4349481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4350963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4352390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4353812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4355277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4356703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4358137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4359558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4360994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4362413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4363817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4365258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4366716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4368241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4369691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4371114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4372532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4373976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4375397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4376825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4378243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4379675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4381080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4382490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4383929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4385406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4386914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4388367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4389802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4391226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4392656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4394076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4395504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4396920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4398335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4399783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4401223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4402644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4404487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4406139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4407626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4409059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4410456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4411882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4413299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4414745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4416155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4417579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4419026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4420473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4421879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4423333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4424879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4426350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4427778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4429343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4430779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4432193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4433604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4435024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4436458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4437872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4439294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4440695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4442120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4443584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4445088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4446557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4447972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4449410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4450813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4457990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4459547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4461013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4462419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4463841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4465250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4466671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4468157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4469713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4471202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4472613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4473998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4475396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4476798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4478213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4479618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4481035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4482449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4483858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4485256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4486695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4488193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4489646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4491065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4492459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4493871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4495282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4496685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4498077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4499477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4500977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4502384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4503793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4505774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4507319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4508760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4510155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4511579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4512990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4514380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4515784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4517187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4518586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4519982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4521393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4522810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4524264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4525758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4527206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4528629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4530049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4531469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4532873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4534291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4535706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4537108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4538509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4539918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4541346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4542776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4544277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4545732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4547147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4548542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4549947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4551347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4552769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 73%] 2024-08-06T21:15:50.4554164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4555572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4556983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4558408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4559810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4561255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4562766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4564226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4565630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4567032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4568447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4569867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4571270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4572670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4574073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4575472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4576867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4578256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4579717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4581219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4582657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4584069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4585475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4586891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4588274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4589698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4591106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4592517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4593907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4595321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4596731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4598193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4599734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4601194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4602607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4604377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4605832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4607236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4608637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4610071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4611465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4612855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4614267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4615675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4617151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4618681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4620166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4621582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4622975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4624379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4625781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4627190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4628577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4629997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4631402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4632817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4634207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4635647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4637138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4638602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4640006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4641401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4642804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4644199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4645585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4646972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4648373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4649794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4651192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4652580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4653977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4655418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4656898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4658327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4659751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4661173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4662565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4663968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4665376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4666794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4668187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4669610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4671018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4672433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4673864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4675357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4676800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4678208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4679609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4681002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4682413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4683823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4685220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4686615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4688030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4689448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4690840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4692281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4693769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4695218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4696612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4698022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4699435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4700891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4702293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4703704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4705474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4706946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4708360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4709773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4711239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4712769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4714207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4715606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4717012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4718438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4719828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4721226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4722638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4724042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4725435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4726831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4728259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4729792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4731277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4732718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4734129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4735537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4736930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4738343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4739756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4741160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4744989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4748230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4749647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4751039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4752545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4754036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4755444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4756847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4758237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4759630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4761028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4762463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4763850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4765242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4766741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4768286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4769678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4771130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4772614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4774030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4775424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4776826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4778234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4779644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4781049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4782472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4783873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4785335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4786795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4788186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4789639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4791097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4792518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4793915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4795329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4796730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4798117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4799519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4800981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4802417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4803857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4805775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4807183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4808696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4810165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4811565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4812989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4814406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4815795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4817183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4818585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4819987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4821374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4822836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4824327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4825723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4827171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4828614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4830009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4831407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4832815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4834204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4835604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4837007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4838402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4839800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4841247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4842732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4844117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4845521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4846974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4848443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4849832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4851237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4852645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4854046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4855426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4856818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4858219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4859661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4861106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4862506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4863912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4865355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4866792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4868180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4869591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4871011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4872402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4873794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4875211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4876614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4878044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4879496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4880911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4882323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4883754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4885210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4886602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4888010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4889396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4890791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4892188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4893598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4894977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4896407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4897872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4899275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4900712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4902149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4903605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4905396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4906816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4908214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4909621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4911397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4912844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4914230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4915626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4917133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4918607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4920004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4921471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4922948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4924327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4925714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4927116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4928573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4929955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4931378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4932777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4934186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4935611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4937057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4938464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4939914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4941389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4942789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4944214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4945624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4947022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4948414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4949825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4951260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4952658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4954100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4955593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4956996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4958424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4959888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4961315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4962740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4964140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4965555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4966957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4968382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4969776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4971210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4972676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4974166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4975559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4976962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4978429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4979889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4981324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4982731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4984162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4985565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4986976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4988378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4989783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4991247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4992710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4994094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4995500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4996962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4998400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.4999845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5001286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5002715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5004453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5005894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5007349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5008772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5010249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5011756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5013169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5014594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5016126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5017593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5019013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5020426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5021861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5023257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5024673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5026084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5027480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5028916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5030392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5031818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5033216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5034649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5036121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5037543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5038933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5040348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5041767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5043183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5044575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5045979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5047428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5048897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5050343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5051769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5053211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5054689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5056078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5057477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5058894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5060288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5061704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5063088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5064508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5065906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5067428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5068886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5070293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5071760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5073219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5074606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5076014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5077444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5078843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5080252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5081684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5083119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5084507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5086025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5087483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5088911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5090305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5091785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5093239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5094644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5096060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5097457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5098875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5100327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5101763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5103152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5105029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5106541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5107948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5109364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5110861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5112357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5113766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5115163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5116576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5118005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5119403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5120828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5122253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5123724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5125170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5126591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5127986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5129444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5130882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5132303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5133705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5135117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5136519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5137915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5139338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5140739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5142167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5143604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5145074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5146476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5147923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5149362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5150780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5152212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5153607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5155014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5156417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5157845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5159238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5160646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5162099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5163565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5164941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5166382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5167832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5169248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5170631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5172059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5173456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5174850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5176250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5177641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5179059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5180509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5182004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5183401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5184865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5186329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5187742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5189141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5190579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5192009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5193403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5194821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5196227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5197640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5199072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5200575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5201999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5203429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5205256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 74%] 2024-08-06T21:15:50.5206756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5208157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5209577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5210971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5212380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5213800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5215214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5216618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5218103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5219581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5221014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5222430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5223871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5225346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5226747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5228155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5229556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5230969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5232381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5233782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5235158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5236611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5238061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5239466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5240848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5242877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5245689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5248312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5250932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5253578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5256222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5258865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5261507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5264370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5267060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5269777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5272473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5275119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5277817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5280534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5283161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5285800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5288445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5291074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5293679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5296307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5298951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5301617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5304735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5307535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5310180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5312870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5315562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5318211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5320860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5323503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5326143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5328767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5331496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5334179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5336828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5339548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5342259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5344904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5347556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5350219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5352902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5355538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5358156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5360796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5363438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5366053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5368675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5371305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5374029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5376710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5379358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5381982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5384667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5387363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5389978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5392628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5395276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5397889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5400558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5403202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5406356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5409017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5411736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5414439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5417048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5419726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5422413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5425033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5427680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5430326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5432923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5435578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5438219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5440846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5443476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5446155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5448862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5451495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5454163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5457569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5460232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5462898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5465518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5468184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5470868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5473523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5476160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5478790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5481476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5484153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5486782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5489420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5492139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5494818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5497443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5500114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5502770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5505855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5508484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5511142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5513797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5516537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5519268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5521928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5524652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5527392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5530111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5532768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5535458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5538130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5540797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5543462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5546129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5548754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5551462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5554176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5556845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5559517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5562217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5564916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5567587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5570252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5572919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5575580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5578289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5580940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5583589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5586314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5589043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5591718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5594371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5597078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5599818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5602520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5605592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5608317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5610974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5613626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5616276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5618936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5621680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5624410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5627060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5629712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5632449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5635191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5637842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5640501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5643174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5645823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5648450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5651113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5653805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5656462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5659182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5661911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5664564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5667267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5669968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5672643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5675300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5677948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5680580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5683214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5685912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5688560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5691222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5693928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5696651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5699278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5702011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5705147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5707862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5710668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5713351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5715999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5718681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5721354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5724010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5726678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5729458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5732182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5734836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5737555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5740322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5742968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5745603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5748265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5750930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5753641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5756300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5758966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5761631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5764331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5767060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5769724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5772445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5775175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5777828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5780497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5783169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5785857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5788511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5791182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5793893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5796554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5799270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5802030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5805092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5807848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5810577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5813243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5815898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5818564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5821246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5823903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5826582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5829248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5831877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5834601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5837340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5840002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5842692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5845411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5848098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5850732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5853386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5856064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5858740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5861401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5864053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5866696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5869400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5872099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5874747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5877407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5880117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5882795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5885427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5888071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5890739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5893386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5896043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5898689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5901422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5904511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5907314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5909991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5913090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5915880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5918639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5921322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5924010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5926692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5929368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5932034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5934711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5937360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5940052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5942764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5945443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5948109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5950788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5953531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5956220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5958881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5961539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5964194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5966844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5969501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5972167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5974886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5977626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5980286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5982939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5985647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5988376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5991056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5993709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5996363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.5999012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6001694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6004684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6007364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6010145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6012864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6015496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6018120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6020886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6023613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6026243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6028929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6031592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6034219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6036890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6039555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6042229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6044884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6047609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6050313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6052987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6055723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6058426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6061078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6063765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6066411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6069038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6071688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6074352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6077007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6079644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6082312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6085037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6087681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6090389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6093092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6095753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6098399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6101061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6103740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6106817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6109484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6112144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6114802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6117557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6120296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6122959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6125673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6128412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6131059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6133675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6136328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6138991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6141644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6144317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6146976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6149622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6152315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6155007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6157656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6160371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6163075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6165697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6168354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6171053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6173718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6176379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6179051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6181729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6184370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6187068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6189782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6192438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6195086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6197788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6200497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6203160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6206252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6208912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6211554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6214230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6216864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6219504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6222229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6224973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6227633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6230272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6233006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6235757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6238412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6241069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6243750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6246422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6249058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6251731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6254384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6257087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6259809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6262461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6265105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6267789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6270508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6273144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6275802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6278487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6281139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6283769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6286429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6289127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6291789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6294488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6297211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6299869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6302620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6305742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6308430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6311121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6313793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6316426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6319090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6321761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6324419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6327074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6329822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6332582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6335216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6337920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6340642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6343311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6345989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6348636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6351269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6353948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6356610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6359278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6361940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6364694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6367397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6370051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6372771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6375517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6378181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6380844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6383515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6386204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6388850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6391495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6394158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 75%] 2024-08-06T21:15:50.6396822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6399501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6402270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6405313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6408095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6410826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6413501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6416160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6418819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6421499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6424133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6426788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6429463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6432114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6434800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6437528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6440237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6442892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6445580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6448296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6450925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6453563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6456192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6458848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6461508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6464157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6466771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6469452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6472163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6474823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6477476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6480173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6482873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6485529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6488184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6490860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6493546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6496230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6498871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6501595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6504728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6507538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6510205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6512863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6515608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6518313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6520960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6523626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6526305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6528990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6531644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6534314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6536988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6539693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6542395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6545062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6547735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6550440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6553167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6555863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6558545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6561212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6563870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6566522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6569208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6571884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6574540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6577246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6580018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6582668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6585361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6588072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6590739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6592148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6593575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6594981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6596414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6597821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6599258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6600701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6602206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6603674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6605476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6606986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6608486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6609902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6611307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6612818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6614234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6615647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6617050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6618468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6619879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6621339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6622802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6624218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6625667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6627131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6628529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6629959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6631376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6632771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6634182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6635597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6637034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6638430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6639899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6641372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6642803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6644207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6645677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6647141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6648575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6650000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6651421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6652846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6654259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6655677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6657070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6658538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6660039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6661457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6662862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6664329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6665791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6667202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6668608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6670062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6671473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6672869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6674287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6675696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6677170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6678615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6680061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6681467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6682937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6684382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6685801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6687199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6688665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6690071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6691463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6692886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6694291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6695735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6697178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6698601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6700025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6701523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6702971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6704762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6706205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6707620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6709015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6710446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6711879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6713278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6714692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6716171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6717676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6719072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6720573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6722042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6723458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6724848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6726260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6727659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6729085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6730506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6731903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6733322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6734787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6736249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6737644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6739107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6740597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6742016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6743414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6744846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6746261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6747673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6749082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6750522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6751952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6753396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6754864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6756272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6757730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6759168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6760597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6761994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6763422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6764820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6766231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6767635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6769064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6770482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6771938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6773412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6774821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6776230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6777665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6779142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6780575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6781986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6783389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6784808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6786221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6787635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6789028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6790485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6791953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6793340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6794739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6796179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6797665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6799046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6800518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6801931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6803346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6805109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6806595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6808001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6809483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6810982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6812383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6813806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6815272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6816757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6818156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6819582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6821020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6822434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6823833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6825249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6826653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6828050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6829480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6830961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6832380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6833815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6835281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6836683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6838110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6839503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6840938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6842343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6843776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6845171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6846582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6848033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6849497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6850933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6852375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6853855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6855265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6856680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6858085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6859504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6860927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6862334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6863730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6865145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6866599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6868046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6869457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6870885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6872367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6873813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6875226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6876637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6878071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6879456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6880900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6882315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6883740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6885229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6886709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6888119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6889537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6891015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6892467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6893881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6895290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6896702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6898093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6899520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6900991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6902398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6903846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6905740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6907161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6908553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6910042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6911535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6912960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6914712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6916161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6917569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6918999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6920393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6921818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6923214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6924719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6926188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6927603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6929042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6930496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6931909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6933297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6934726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6936126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6937532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6938922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6940337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6941752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6943196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6945423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6946849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6948307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6949757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6951173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6952604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6954044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6955446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6956867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6958278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6959720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6961112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6962590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6964047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6965465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6966889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6968336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6969758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6971168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6972596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6973994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6975413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6976832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6978252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6979647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6981113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6982592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6984004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6985401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6986869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6988339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6989738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6991162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6992583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6994009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6995410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6996840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6998237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.6999705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7001194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7002611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7004401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7005953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7007427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7008824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7010253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7011683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7013098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7014499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7015923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7017327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7018802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7020264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7021707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7023116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7024568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7026026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7027431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7028860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7030259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7031705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7033105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7034532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7035916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7037319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7038763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7040235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7041633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7043087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7044537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7045957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7047350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7048745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7050169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7051604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7053029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7054433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7055862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7057320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7058789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7060195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7061676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7063146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7064561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7065965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7067374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7068800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7070184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7071607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7073018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7074448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7075881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7077346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7078752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7080226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7081700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7083120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7084514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7095173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7096729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7098146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7099559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7101066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7102471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7104286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7105917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7107366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7108774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7110239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7111713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7113105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7114499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7115883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7117285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7118688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7120091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7121479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7122926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7124398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7125784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7127171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7128607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7130078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7131471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7132877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7134282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7135696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7137088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7138499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7139895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7141307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7142754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7144201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7145598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7147036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 76%] 2024-08-06T21:15:50.7148485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7149864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7151286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7152694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7154087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7155468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7156880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7158275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7159665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7161096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7162544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7163957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7165400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7166858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7168250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7169668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7171068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7172468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7173863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7175278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7176673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7178078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7179512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7180970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7182361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7183786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7185303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7186697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7188093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7189478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7190897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7192296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7193693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7195079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7196478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7197919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7199351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7200792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7202194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7203649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7205467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7206882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7208282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7209694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7211085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7212485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7213876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7215273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7216739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7218191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7219599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7221003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7222457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7223916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7225318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7226715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7228115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7229496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7230914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7232329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7233728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7235115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7236571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7238042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7239425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7240876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7242330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7243751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7245143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7246553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7247943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7249346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7250726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7252123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7253515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7254970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7256407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7257791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7259236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7260694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7262086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7263470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7264909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7266332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7267749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7269158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7270604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7272026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7273493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7274958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7276371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7277853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7279311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7280756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7282163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7283595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7285000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7286408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7287826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7289264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7290699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7292164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7293629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7295056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7296456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7297905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7299383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7300864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7302290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7303707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7305539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7306998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7308431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7309839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7311390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7312897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7314319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7315733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7317222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7318700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7320101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7321549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7322978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7324409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7325815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7327233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7328642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7330120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7331602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7333024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7334440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7335913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7337370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7338779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7340219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7341669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7343085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7344492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7345931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7347345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7348799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7350258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7351704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7353112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7354549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7356003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7357421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7358841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7360244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7361677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7363085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7364516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7365911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7367367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7368838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7370279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7371703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7373161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7374652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7376088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7377505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7378925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7380357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7381800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7383227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7384637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7386072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7387530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7388999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7390399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7391888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7393364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7394780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7396192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7397625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7399044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7400486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7401920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7403343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7405173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7406725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7408236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7409660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7411161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7412648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7414071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7415492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7416933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7418340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7419754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7421191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7422632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7424043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7425492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7426973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7428393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7429846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7431330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7432763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7434182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7435598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7437001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7438417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7439833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7441252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7442669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7444125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7445604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7447004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7448482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7449951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7451396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7452799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7454221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7455627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7457046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7458442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7459842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7461275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7462732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7464195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7465594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7467054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7468520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7469917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7471337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7472767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7474188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7475600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7477009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7478435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7479857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7481329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7482806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7484226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7485665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7487185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7488672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7490083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7491539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7492942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7494362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7495780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7497215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7498615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7500062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7501603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7503024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7504836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7506358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7507843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7509260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7510676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7512103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7513530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7514955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7516377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7517776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7519260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7520770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7522185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7523606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7525060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7526536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7527929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7529341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7530756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7532197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7533596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7535011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7536416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7537882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7539329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7540734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7542171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7543626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7545085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7546488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7547918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7549340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7550753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7552178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7553605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7555016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7556472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7557928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7559349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7560756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7562188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7563649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7565055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7566488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7567933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7569347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7570749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7572169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7573566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7574975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7576431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7577914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7579312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7580764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7582238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7583655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7585073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7586487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7587912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7589332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7590755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7592163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7593573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7595056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7596526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7597925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7599395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7600923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7602337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7603765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7605562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7606995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7608389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7609822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7611254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7612689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7614169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7615662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7617072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7618564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7620034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7621456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7622887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7624311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7625728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7627132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7628555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7629971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7631393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7632838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7634313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7635722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7637133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7638569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7640039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7641472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7642874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7644294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7645705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7647141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7648548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7649971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7651455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7652949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7654349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7655777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7657236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7658737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7660141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7661574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7663005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7664419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7665832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7667233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7668656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7670114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7671605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7673006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7674425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7675883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7677356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7678755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_1024_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7680186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7681635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7683039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7684462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7685882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7687315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7688760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7690231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7691670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7693108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7694556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7696075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7697481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7698908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7700306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7701820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7703253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7705051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7706534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7708022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7709524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7710940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7712374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7713835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7715330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7716749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7718169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7719582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7721017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7722463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7723873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7725303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7726718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7728204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7729658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7731080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7732560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7734045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7735438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7736861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7738276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7739711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7741108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7742534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7743966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7745382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7746844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7748309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7749736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7751189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7752670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7754064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7755492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7756911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7758324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7759725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7761155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7762589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7763984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7765451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7766905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7768326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7769756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7771213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7772632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7774055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7775453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7776866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7778262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7779693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7781091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7782500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7783976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7785521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7786936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7788342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7789848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7791333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7792768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7794181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7795623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7797037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7798462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7799879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 77%] 2024-08-06T21:15:50.7801342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7802810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7804775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7806227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7807644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7809165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7810629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7812061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7813473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7814904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7816297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7817717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7819142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7820570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7822056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7823553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7824990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7826404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7827867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7829339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7830773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7832208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7833638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7835047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7836478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7837896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7839309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7840749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7842248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7843667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7845063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7846526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7847993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7849426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7850834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7852277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7853680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7855107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7856506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7857923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7859330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7860806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7862274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7863678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7865149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7866620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7868035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7869439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7870862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7872287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7873696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7875088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7876517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7877921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7879386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7880831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7882251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7883720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7885165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7886577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7887990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7889430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7890832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7892275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7893696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7895136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7896535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7898003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7899470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7900948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7902424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7903900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7905782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7907202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7908621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7910021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7911446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7912884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7914307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7916072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7917630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7919130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7920544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7921951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7923444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7924939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7926330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7927753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7929169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7930596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7931994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7933446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7934852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7936330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7937783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7939204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7940609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7942071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7943545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7944935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7946360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7947776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7949179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7950578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7952003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7953429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7954879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7956322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7957750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7959161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7960617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7962068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7963502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7964937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7966340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7967753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7969167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7970606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7972005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7973484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7974934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7976364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7977749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7979202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7980662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7982090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7983506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7984909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7986330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7987741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7989155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7990559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7991984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7993460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7994924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7996318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7997796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.7999263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8000719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8002142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8003583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8005427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8006897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8008329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8009732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8011154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8012629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8014134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8015540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8017039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8018504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8019912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8021317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8022730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8024153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8025544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8026974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8028384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8029788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8031223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8032698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8034122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8035568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8037014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8038437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8039841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8041254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8042652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8044064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8045486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8046882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8048282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8049724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8051195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8052578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8054006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8055449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8056919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8058298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8059709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8061123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8062532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8063963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8065363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8066787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8068242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8069706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8071110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8072534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8074008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8075475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8076874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8078294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8079703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8081091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8082504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8083926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8085354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8086743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8088254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8089731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8091153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8092587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8094064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8095480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8096942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8098360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8099793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8101259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8102699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8104552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8106009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8107534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8109043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8110475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8111949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8113463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8114880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8116299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8117708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8119142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8120571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8121985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8123425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8124844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8126331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8127790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8129217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8130678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8132174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8133590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8135030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8136461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8137907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8139318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8140754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8142191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8143637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8145117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8146589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8148026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8149486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8150959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8152366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8153828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8155255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8156683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8158090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8159529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8160957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8162379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8163869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8165344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8166784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8168235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8169722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8171139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8172581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8174017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8175452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8176869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8178310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8179718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8181141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8182589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8184079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8185495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8186897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8188370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8189838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8191255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8192668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8194129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8195547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8196967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8198377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8199820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8201330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8202811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8204609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8206073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8207604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8209088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8210522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8211952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8213403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8214833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8216272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8217693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8219126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8220584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8222080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8223502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8224944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8226411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8227890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8229327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8230756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8232189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8233601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8235061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8236494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8237921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8239379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8240877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8242308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8243734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8245205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8246679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8248124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8249540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8250984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8252398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8253836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8255246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8256670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8258125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8259614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8261022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8262447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8263923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8265400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8266822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8268236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8269673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8271091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8272517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8273936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8275377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8276839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8278315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8279728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8281160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8282620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8284104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8285512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8286920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8288354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8289746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8291161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8292576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8294027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8295426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8296895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8298361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8299794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8301287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8302777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8304577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8306040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8307510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8308927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8310375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8311809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8313243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8314680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8316206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8317705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8319139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8320613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8322119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8323541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8324991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8326409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8327833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8329281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8330696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8332123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8333539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8335053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8336512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8337937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8339396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8340879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8342283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8343709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8345159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8346604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8348010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8349426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8350868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8352289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8353757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8355240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8356665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8358125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8359594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8360995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8362432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8363854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8365301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8366705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8368125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8369568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8370968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8372437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8373924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8375389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8376845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8378381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8379798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8381238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8382647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8384080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8385523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8386969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8388379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8389787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8391258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8392727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8394140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8395564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8397038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8398501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8399916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8401360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8402801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8404580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8406052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8407462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8408873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8410390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8411865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8413289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8414709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8416221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8417697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8419116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8420534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8421980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8423388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8424819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8426244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8427671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8429107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8430554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8431987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8433400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8434861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8436325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8437752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8439175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8440602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8442005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8443437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8444862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8446278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8447737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8449206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8450645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8452038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8453502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8454993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 78%] 2024-08-06T21:15:50.8456431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8457833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8459260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8460661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8462084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8463476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8464892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8466302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8467760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8469223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8470623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8472089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8473553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8474970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8476377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8477812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8479233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8480658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8482064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8483507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8484939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8486405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8487870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8489290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8490773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8492242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8493673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8495101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8496538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8497936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8499355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8500803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8502247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8503647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8505539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8507047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8508464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8509942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8511433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_2048_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8512865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8514284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8515734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8517144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8518584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8520006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8521432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8522842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8524324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8525818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8527243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8528654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8530106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8531586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8532980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8534406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8535846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8537282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8538683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8540107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8541512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8542988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8544431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8545876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8547294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8548754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8550221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8551628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8553071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8554495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8555937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8557352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8558788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8560205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8561676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8563138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8564564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8565992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8567451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8568907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8570317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8571753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8573160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8574580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8576019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8577456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8578856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8580316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8581772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8583209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8584606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8586105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8587573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8588989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8590409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8591818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8593244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8594665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8596109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8597508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8598923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8600376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8601885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8603284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8605163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8606735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8608149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8609549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8610961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8612389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8613783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8615210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8616653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8618081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8619539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8621037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8622457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8623938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8625399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8626843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8628263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8629687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8631109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8632516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8633949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8635369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8636798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8638241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8639724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8641134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8642591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8644045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8645472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8646900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8648322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8649729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8651137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8652574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8653975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8655394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8656867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8658368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8659773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8661235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8662699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8664133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8665975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8667458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8668890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8670304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8671736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8673145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8674571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8676056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8677530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8678933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8680364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8681875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8683339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8684737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8686193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8687610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8689021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8690422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8691838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8693274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8694716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8696191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8697601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8699038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8700483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8701982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8703380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8705228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8706656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8708062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8709466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8710880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8712286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8713766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8715254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8716675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8718085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8719538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8721037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8722457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8723880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8725289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8726739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8728165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8729591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8730999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8732473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8733970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8735381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8736829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8738280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8739767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8741157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8742569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8743983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8745412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8746831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8748258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8749671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8751085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8752547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8754001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8755424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8756927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8758394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8759793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8761225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8762641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8764050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8765449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8766903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8768318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8769726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8771170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8772619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8774036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8775463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8776945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8778348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8779776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8781171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8782590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8783993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8785440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8786856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8788298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8789754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8791239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8792635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8794075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8795572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8797012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8798431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8799840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8801395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8802817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8804684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8806124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8807564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8809049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8810524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8811913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8813337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8815148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8817764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8819245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8820660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8822096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8823485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8824902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8827163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8829151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8830669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8832147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8833566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8835195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8836874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8839205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8842224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8844221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8845667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8847344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8849737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8852245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8854885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8858016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8859563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8861623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8863054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8864699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8867154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8868613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8871432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8873088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8875770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8877440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8879526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8881224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8882656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8884581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8886987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8888830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8890513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8892027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8893821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8895613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8897281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8898726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8900132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8901558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8902986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8904856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8906450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8907935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8909374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8910839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8912310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8913703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8915113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8916550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8918322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8919780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8921197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8922642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8924036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8925522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8926989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8928422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8929840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8931301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8932747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8934152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8935557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8936953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8938378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8939812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8941227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8942681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8944143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8945601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8947009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8948405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8949892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8951343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8952748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8954145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8955555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8956981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8958379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8959818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8961226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8962693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8964153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8965572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8966969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8968504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8969962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8971363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8972769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8974182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8975593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8976987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8978408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8979839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8981244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8982676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8984152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8985557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8987014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8988456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8989899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8991315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8992710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8994155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8995563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8996995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8998396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.8999832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9001314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9002854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9004619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9006145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9007689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9009110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9010522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9011936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9013343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9014745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9016154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9017550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9018967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9020461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9021935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9023326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9024751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9026221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9027679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9029078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9030517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9031929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9033316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9034734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9036134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9037550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9038984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9040460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9041856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9043269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9044698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9046153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9047546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9048969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9050380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9051762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9053189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9054601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9056007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9057452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9058940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9060377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9061792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9063235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9064718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9066126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9067543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9068945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9070364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9071783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9073173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9074587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9075987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9077456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9078898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9080326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9081775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9083631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9085054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9086470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9087881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9089312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9090741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9092177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9093624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9095081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9096564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9098021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9099449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9100971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9102467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9103870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9105720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9107134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9108540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9109927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9111370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9112786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9114171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9115667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9117139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9118561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9119947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9121441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9122912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9124331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9125788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9127202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9128599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9130022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9131435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 79%] 2024-08-06T21:15:50.9132825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9134294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9135751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9137157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9138546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9139998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9141467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9142859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9144244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9145664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9147062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9148463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9149851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9151257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9152676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9154104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9155554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9156956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9158419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9159860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9161290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9162696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9164128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9165522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9166940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9168348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9169781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9171196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9172637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9174105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9175506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9176959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9178399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9179813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9181240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9182649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9184044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9185457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9186866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9188277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9189665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9191132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9192599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9193978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9195378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9196819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9198288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9199676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9201144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9202553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9204313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9205820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9207277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9208677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9210158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9211645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9213024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9214434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9215889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9217357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9218743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9220160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9221588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9222982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9224370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9225786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9227190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9228618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9230074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9231505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9232933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9234368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9235834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9237241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9238691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9240093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9241556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9242952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9244372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9245755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9247152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9248598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9250064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9251482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9252978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9254446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9255847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9257251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9258638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9260635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9262604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9264316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9266361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9268796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9271093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9273021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9275064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9277024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9279145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9281006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9282951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9284656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9286581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9288147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9290023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9292089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9293996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9295946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9297754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9299637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9301427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9303098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9305169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9307112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9308547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9310248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9311985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9313846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9315679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9317498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9319300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9321238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9323045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9324821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9326568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9328465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9330157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9331920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9333768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9335576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9337235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9338883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9340676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9342223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9343799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9345243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9346623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9348080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9349535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9350946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9352337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9353763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9355173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9356578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9357974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9359396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9360815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9362374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9363818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9365210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9367197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9368689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9370167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9371585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9373012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9374645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9376073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9377473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9378895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9380273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9381732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9383264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9384709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9386106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9387569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9389034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9390472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9391886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9393299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9394733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9396153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9397577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9398978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9400442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9401959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9403381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9405171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9406716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9408227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9409636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9411039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9412472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9413900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9415298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9416722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9418201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9419637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9421097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9422606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9424026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9425505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9426961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9428384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9429800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9431243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9432662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9434074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9435499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9436910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9438316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9439758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9441236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9442669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9444186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9445646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9447129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9448553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9449973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9451371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9452805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9454235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9455634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9457045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9458501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9459983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9461377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9462860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9464379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9465810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9467206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9468632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9470032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9471451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9472869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9474268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9475745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9477207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9478667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9480059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9481540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9483022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9484436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9485828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9487267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9488687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9490108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9491516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9493019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9494464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9495915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9497388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9498804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9500246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9501731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9503229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9505039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9506538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9507943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9509355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9510763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9512197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9513611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9515092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9516593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9518007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9519421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9520878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9522719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9524179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9525600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9527010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9528449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9529886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9531368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9532776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9534267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9535756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9537162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9538586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9540086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9541574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9542967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9544401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9545816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9547241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9548640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9550063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9551469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9552948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9554421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9555826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9557250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9558704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9560165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9561556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9562985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9564417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9565830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9567231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9568660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9570075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9571490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9572933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9574391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9575807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9577232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9578752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9580158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9581588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9582985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9584397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9585792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9587215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9588609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9590018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9591473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9592976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9594380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9595846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9597315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9598733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9600160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9601614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9603076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9604869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9606330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9607808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9609233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9610725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9612209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9613610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9615103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9616660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9618055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9619475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9620892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9622384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9623794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9625221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9626634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9628057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9629508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9630971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9632381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9633827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9635268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9636734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9638203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9639764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9641212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9642732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9644199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9645664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9647060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9648528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9650039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9651471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9652865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9654340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9655791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9657253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9658670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9660072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9661494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9662911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9664350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9665748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9667219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9668684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9670101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9671499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9672970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9674460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9675860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9677337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9678748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9680192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9681626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9683026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9684487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9685897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9687348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9688790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9690192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9691659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9693113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9694546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9695956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9697445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9698838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9700279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9701727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9703165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9704959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9706503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9707991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9709409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9710887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9712360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9713783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9715220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9716632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9718022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9719441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9720854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9722267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9723661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9725154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9726618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9728031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9729470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9730927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9732349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9733737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9735178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9736589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9738015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9739410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9740832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9742234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9743706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9745238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9746659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9748053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9749491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9750940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9752323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9753743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9755177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9756585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9757981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9759400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9760800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9762248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9763681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9765127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9766536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9767985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9769451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9770862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9772296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9773704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9775147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9776564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9778000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9779396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9780855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9782300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9783719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9785127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9786577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9788028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9789435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9790846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9792243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9793662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9795097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9796512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9797911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_587_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9799333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9800789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9802287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9803685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 80%] 2024-08-06T21:15:50.9805577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9807131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9808538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9809960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9811378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9812809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9814210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9815664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9817071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9818492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9819937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9821410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9822814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9824281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9825749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9827164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9828574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9829992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9831403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9832802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9834237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9835676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9837156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9838603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9840103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9841575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9843045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9844498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9845946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9847361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9848762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9850190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9851591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9853022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9854412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9855850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9857301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9858779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9860173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9861584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9863027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9865261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9866674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9868078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9869508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9870912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9872321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9873721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9875144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9876620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9878082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9879475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9880894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9882342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9883806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9885200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9886640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9888050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9889492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9890906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9892315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9893735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9895163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9903681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9905547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9906999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9908505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9909998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9911409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9912847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9914264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9915676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9917089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9918799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9920345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9921766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9923288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9924804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9926225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9927792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9929279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9930687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9932101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9933499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9934937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9936366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9937786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9939185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9940605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9942060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9943501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9944929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9946379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9947858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9949248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9950666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9952080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9953502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9954909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9956327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9957739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9959163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9960603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9962066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9963480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9964943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9966398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9967785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9969203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9970614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9972027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9973415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9974857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9976269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9977671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9979106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9980571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9981972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9983355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9984821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9986284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9987700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9989095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9990511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9991911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9993413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9994824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9996235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9997669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:50.9999131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0000516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0001933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0003390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0005316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0006745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0008138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0009555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0010957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0012357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0013747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0015187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0016598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0018091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0019562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0020986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0022462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0023923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0025353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0026768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0028207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0029607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0031029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0032435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0033863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0035248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0036698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0038154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0039582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0041019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0042471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0043889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0045304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0046719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0048117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0049532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0050938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0052348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0053739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0055206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0056662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0058141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0059602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0061063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0062488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0063938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0065365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0066758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0068173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0069559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0071026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0072425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0073891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0075328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0076731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0078128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0079586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0081018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0082404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0083826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0085240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0086640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0088035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0089461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0090871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0092324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0093761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0095186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0096590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0098038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0099481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0100931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0102382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0103769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0105557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0107015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0108441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0109824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0111309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0112777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0114202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0115587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0117049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0118581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0119986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0121393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0122789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0124207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0125616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0127031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0128428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0129847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0131303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0132779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0134181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0135633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0137090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0138476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0139872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0141274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0142696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0144081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0145491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0146891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0148306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0149731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0151178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0152567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0154021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0155453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0156856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0158249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0159658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0161066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0162453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0163871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0165280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0166679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0168107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0169556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0170945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0172332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0173754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0175213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0176604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0177982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0179390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0180776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0182191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0183578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0184994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0186435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0187904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0189289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0190693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0192138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0193612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0195012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0196407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0197826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0199231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0200639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0202076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0203493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0205264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0206779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0208239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0209652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0211119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0212589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0213974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0215400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0216808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0218192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0219591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0220997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0222427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0223810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0225336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0226855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0228281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0229716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0231188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0232588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0234014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0235425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0236824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0238239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0239644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0241047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0242437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0243893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0245358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0246760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0248149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0249602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0251051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0252455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0253848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_128_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0255273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0256690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0258082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0259495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0260893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0262360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0263800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0265225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0266630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0268658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0271522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0274169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0276825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0279516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0282188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0284820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0287471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0290107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0292842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0295525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0298188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0300849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0303550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0306679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_160_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0309322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0311973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0314608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0317224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0319860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0322515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0325174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0327796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0330533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0333288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0335900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0338600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0341303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0343933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0346557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0349167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0351774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0354405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0357044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0359690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0362309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0364994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0367652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0370266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0372938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0375665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0378305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0380930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0383561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0386248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0388897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0391531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0394222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0396882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0399559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0402287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0405311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0408031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0410758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0413466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0416081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0418732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0421402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0424031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0426678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0429348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0431994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0434661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_192_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0437372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0440027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0442667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0445339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0448059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0450717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0453379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0456046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0458697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0461369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0464029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0466648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0469296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0471993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0474681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0477299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0479977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0482681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0485315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0487958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0490600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0493251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0495896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0498510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_203_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0501146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0503829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0506996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0509714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0512342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0515030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0517741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0520363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0522997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0525674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0528318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0530924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0533557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0536218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0538833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0541488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0544223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0546861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0549466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0552132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0554811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0557448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0560079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0562719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_21_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0565338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0567997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0570670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0573298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0575991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0578725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0581373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0584014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0586716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0589437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0592084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0594725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0597357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0600019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0602686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0605725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0608366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0611092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0613784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0616411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0619061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0621777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0625768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0630376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_256_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0635457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0641050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 81%] 2024-08-06T21:15:51.0646537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0652125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0657732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0663334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0668877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0674571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0680231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0685827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0691549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0697161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0702810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0708856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0714427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0719917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0725437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0731005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0736531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0742001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0747734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0753400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0758917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0764578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0770256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0775740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0781315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0786847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0792379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0797954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0803569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0809421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0814980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0820767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0826449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0831977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0837494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0843252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0848899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0854389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0859956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0865527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0871056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0876569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0882102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0887676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0893406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0899004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0904201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0908832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0914278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0919338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0924503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0928412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0931075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0933674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0936310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0938957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0941586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0944224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0947014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0949711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0952318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0954966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0957657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0960290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0962928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0965522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0968143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0970767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0973390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0976001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_72_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0978611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0981327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0983975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0986579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0989241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0991931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0994546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0997152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.0999781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1002486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1005590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1008303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1010934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1013552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1016254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1018960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1021582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1024210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1026887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1029569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1032162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1034796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1037444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1040048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1042695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1045340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1047947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1050626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1053313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1055991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1058632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1061301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1063974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1066624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1069276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1071902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_False_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1074548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1077208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1079815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1082424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1085049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1087731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1090408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1093042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1095708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale0_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1098395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_bfloat16_scale_l1_cuda_bfloat16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1101021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1103696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_96_is_causal_True_dropout_p_0_48_float16_scale_l1_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1106912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1109792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1112692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1115550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1118422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1121370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1124342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1127220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1130087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1132990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1135941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1138799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1141661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1144532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1147389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1150277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1153122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1155979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1158887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1161792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1164646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1167574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1170485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1173317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1176173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1179022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1181879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1184739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1187620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1190467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1193340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1196247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1199126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1202064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1205386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1208268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1211140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1213977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1216832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1219692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1222541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1225405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1228323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1231281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1234109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1236952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1239857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1242808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1245655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1248498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1251337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1254180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1257031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1259884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1262766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1265678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1268569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1271378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1274281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1277170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1280014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1282856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1285708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1288543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_32_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1291379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1294242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1297104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1300036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1302987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1306319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1309312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1312228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1315083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1317969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1320840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1323700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1326563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1329417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1332253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1335170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1338141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1340984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1343836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1346733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1349644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1352476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1355348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1358205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1361044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1363883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1366739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1369580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1372479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1375390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1378233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1381138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_256_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1384042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1386888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1389722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1392582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1395430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1398269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1401167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1404427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1407408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1410305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1413206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1416132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1419050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1421882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1424737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1427598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_256_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1430436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1433300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1436155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1439003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1441895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1444807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1447652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1450510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_64_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1453391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1456317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1459156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1462002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_0_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1464840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1467667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale0_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1470516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_False_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1473364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_nestedtensor_batch_size_8_max_seq_len_q_32_max_seq_len_kv_32_head_dim_8_dropout_p_0_1_float16_scale_l1_is_causal_True_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1476194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_32_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1479042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_32_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1482573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_32_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1485365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_32_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1488212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_64_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1491078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_64_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1493866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_64_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1496677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths0_head_dim_64_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1499476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_32_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1502315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_32_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1505504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_32_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1508342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_32_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1511137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_64_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1513933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_64_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1516803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_64_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1519664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths1_head_dim_64_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1522448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_32_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1525304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_32_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1528181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_32_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1530964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_32_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1533763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_64_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1536563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_64_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1539348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_64_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1542118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_sequence_legnths2_head_dim_64_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1544942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_32_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1547762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_32_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1550552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_32_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1553396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_32_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1556241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_64_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1559024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_64_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1561858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_64_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1564703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths0_head_dim_64_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1567497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_32_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1570302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_32_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1573127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_32_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1575915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_32_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1578695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_64_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1581505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_64_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1584315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_64_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1587150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths1_head_dim_64_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1589989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_32_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1592810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_32_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1595581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_32_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1598418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_32_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1601265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_64_is_causal_False_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1604537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_64_is_causal_False_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1607411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_64_is_causal_True_dropout_p_0_0_float16_fused_kernel0_cuda_float16 SKIPPED [0.0002s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1610213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_sequence_legnths2_head_dim_64_is_causal_True_dropout_p_0_22_float16_fused_kernel0_cuda_float16 SKIPPED [0.0003s] (Does not support SDPA or pre-SM80 hardware) [ 82%] 2024-08-06T21:15:51.1612416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_backwards_throws_determinism_warning_fused_kernel0_warn_only_False_cuda PASSED [0.0042s] [ 82%] 2024-08-06T21:15:51.1614157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_backwards_throws_determinism_warning_fused_kernel0_warn_only_True_cuda PASSED [0.0074s] [ 82%] 2024-08-06T21:15:51.1616466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0234s] [ 82%] 2024-08-06T21:15:51.1619202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0226s] [ 82%] 2024-08-06T21:15:51.1621918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0224s] [ 82%] 2024-08-06T21:15:51.1624724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0233s] [ 82%] 2024-08-06T21:15:51.1627506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0225s] [ 82%] 2024-08-06T21:15:51.1630215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0231s] [ 82%] 2024-08-06T21:15:51.1632913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0232s] [ 82%] 2024-08-06T21:15:51.1635679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0188s] [ 82%] 2024-08-06T21:15:51.1638456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0210s] [ 82%] 2024-08-06T21:15:51.1641142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0206s] [ 82%] 2024-08-06T21:15:51.1643871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0215s] [ 82%] 2024-08-06T21:15:51.1646574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0215s] [ 82%] 2024-08-06T21:15:51.1649263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0217s] [ 82%] 2024-08-06T21:15:51.1651973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0214s] [ 82%] 2024-08-06T21:15:51.1654673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0216s] [ 82%] 2024-08-06T21:15:51.1657342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0179s] [ 82%] 2024-08-06T21:15:51.1660062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0198s] [ 82%] 2024-08-06T21:15:51.1662831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0215s] [ 82%] 2024-08-06T21:15:51.1665589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0209s] [ 82%] 2024-08-06T21:15:51.1668307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0224s] [ 82%] 2024-08-06T21:15:51.1671049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0208s] [ 82%] 2024-08-06T21:15:51.1673783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0219s] [ 82%] 2024-08-06T21:15:51.1676476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0217s] [ 82%] 2024-08-06T21:15:51.1679159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0175s] [ 82%] 2024-08-06T21:15:51.1681857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0227s] [ 82%] 2024-08-06T21:15:51.1684561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0205s] [ 82%] 2024-08-06T21:15:51.1687249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0196s] [ 82%] 2024-08-06T21:15:51.1689925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0216s] [ 82%] 2024-08-06T21:15:51.1692621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0200s] [ 82%] 2024-08-06T21:15:51.1695313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0200s] [ 82%] 2024-08-06T21:15:51.1698048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0208s] [ 82%] 2024-08-06T21:15:51.1700780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_False_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0162s] [ 82%] 2024-08-06T21:15:51.1703541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0200s] [ 82%] 2024-08-06T21:15:51.1706676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0199s] [ 82%] 2024-08-06T21:15:51.1709475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0199s] [ 82%] 2024-08-06T21:15:51.1712273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0214s] [ 82%] 2024-08-06T21:15:51.1714971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0198s] [ 82%] 2024-08-06T21:15:51.1717663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0208s] [ 82%] 2024-08-06T21:15:51.1720380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0205s] [ 82%] 2024-08-06T21:15:51.1723067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0171s] [ 82%] 2024-08-06T21:15:51.1725759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0190s] [ 82%] 2024-08-06T21:15:51.1728467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0195s] [ 82%] 2024-08-06T21:15:51.1731143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0200s] [ 82%] 2024-08-06T21:15:51.1733828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0203s] [ 82%] 2024-08-06T21:15:51.1736570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0200s] [ 82%] 2024-08-06T21:15:51.1739295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0204s] [ 82%] 2024-08-06T21:15:51.1741995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0202s] [ 82%] 2024-08-06T21:15:51.1744677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_False_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0163s] [ 82%] 2024-08-06T21:15:51.1747409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0184s] [ 82%] 2024-08-06T21:15:51.1750153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0193s] [ 82%] 2024-08-06T21:15:51.1752839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0197s] [ 82%] 2024-08-06T21:15:51.1755501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0202s] [ 82%] 2024-08-06T21:15:51.1758195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0197s] [ 82%] 2024-08-06T21:15:51.1760882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0196s] [ 82%] 2024-08-06T21:15:51.1763592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0198s] [ 82%] 2024-08-06T21:15:51.1766273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_False_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0164s] [ 82%] 2024-08-06T21:15:51.1768948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0034s] [ 82%] 2024-08-06T21:15:51.1771680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0034s] [ 82%] 2024-08-06T21:15:51.1774433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0035s] [ 82%] 2024-08-06T21:15:51.1777106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_False_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0034s] [ 82%] 2024-08-06T21:15:51.1779796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_False_cuda PASSED [0.0034s] [ 82%] 2024-08-06T21:15:51.1782521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_False_expand_v_num_heads_True_cuda PASSED [0.0034s] [ 82%] 2024-08-06T21:15:51.1785246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_False_cuda PASSED [0.0037s] [ 82%] 2024-08-06T21:15:51.1787896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_kernel0_expand_q_batch_True_expand_k_batch_True_expand_v_batch_True_expand_q_num_heads_True_expand_k_num_heads_True_expand_v_num_heads_True_cuda PASSED [0.0034s] [ 82%] 2024-08-06T21:15:51.1789907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_nested_broadcasting_query_dense_cuda PASSED [0.0283s] [ 82%] 2024-08-06T21:15:51.1791376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_kernels_seq_len_1_inputs_fused_kernel0_cuda PASSED [0.0282s] [ 82%] 2024-08-06T21:15:51.1792748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_sdp_choice_type_dense_cuda PASSED [0.0017s] [ 82%] 2024-08-06T21:15:51.1794026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_fused_sdp_choice_type_nested_cuda PASSED [0.0019s] [ 82%] 2024-08-06T21:15:51.1795452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_attention_long_sequence_mask_float16_cuda_float16 PASSED [0.0012s] [ 82%] 2024-08-06T21:15:51.1796231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_attention_long_sequence_mask_float32_cuda_float32 PASSED [0.0012s] [ 82%] 2024-08-06T21:15:51.1796952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_attention_non_contig_mask_bug_cuda PASSED [0.0017s] [ 82%] 2024-08-06T21:15:51.1797746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_attention_non_contiguous_mask_float16_cuda_float16 PASSED [0.0020s] [ 82%] 2024-08-06T21:15:51.1798556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_attention_non_contiguous_mask_float32_cuda_float32 PASSED [0.0033s] [ 82%] 2024-08-06T21:15:51.1799278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_attention_pad_mask_float16_cuda_float16 PASSED [0.0019s] [ 82%] 2024-08-06T21:15:51.1800000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_attention_pad_mask_float32_cuda_float32 PASSED [0.0019s] [ 82%] 2024-08-06T21:15:51.1801090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_eff_backwards_determinism_cuda SKIPPED [0.0002s] (This test is not behaving deterministaclly non-deterministaclly on CI/CD) [ 82%] 2024-08-06T21:15:51.1802462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0188s] [ 82%] 2024-08-06T21:15:51.1803823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 82%] 2024-08-06T21:15:51.1805492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0081s] [ 82%] 2024-08-06T21:15:51.1808462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 SKIPPED [0.0006s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131105 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1811381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0006s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131126 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1814210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0006s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131168 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1815611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 [W806 21:12:58.568429971 attention.cpp:807] Warning: Dropout mask should only be used for testing purposes. (function operator()) 2024-08-06T21:15:51.1815860Z PASSED [0.0078s] [ 82%] 2024-08-06T21:15:51.1818668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 SKIPPED [0.0006s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131383 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1819996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0066s] [ 82%] 2024-08-06T21:15:51.1822853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 SKIPPED [0.0006s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131435 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1825736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 SKIPPED [0.0006s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131090 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1828598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 SKIPPED [0.0007s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131139 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1831487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0006s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131495 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1834320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0008s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131609 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1837130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 SKIPPED [0.0006s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131624 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1839956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 SKIPPED [0.0006s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131537 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1842839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 SKIPPED [0.0006s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131181 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1844179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0078s] [ 82%] 2024-08-06T21:15:51.1846990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 SKIPPED [0.0006s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131605 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1849833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 SKIPPED [0.0006s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131606 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1852725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 SKIPPED [0.0006s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131602 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1855529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 SKIPPED [0.0007s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131633 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1858349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 SKIPPED [0.0006s] (Test is disabled because an issue exists disabling it: https://github.com/pytorch/pytorch/issues/131646 for platform(s) linux. If you're seeing this on your local machine and would like to enable this test, please make sure CI is not set and you are not using the flag --import-disabled-tests.) [ 82%] 2024-08-06T21:15:51.1859642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 82%] 2024-08-06T21:15:51.1860949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 82%] 2024-08-06T21:15:51.1862286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 82%] 2024-08-06T21:15:51.1863626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 82%] 2024-08-06T21:15:51.1864902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 82%] 2024-08-06T21:15:51.1866196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0099s] [ 82%] 2024-08-06T21:15:51.1867516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0088s] [ 82%] 2024-08-06T21:15:51.1868916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 ('RERUN', {'yellow': True}) [0.0054s] [ 82%] 2024-08-06T21:15:51.1870247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 ('RERUN', {'yellow': True}) [0.0060s] [ 82%] 2024-08-06T21:15:51.1871562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 82%] 2024-08-06T21:15:51.1872851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 82%] 2024-08-06T21:15:51.1874174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 ('RERUN', {'yellow': True}) [0.0085s] [ 82%] 2024-08-06T21:15:51.1875514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 ('RERUN', {'yellow': True}) [0.0085s] [ 82%] 2024-08-06T21:15:51.1876790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0094s] [ 82%] 2024-08-06T21:15:51.1878093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0099s] [ 82%] 2024-08-06T21:15:51.1879420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 ('RERUN', {'yellow': True}) [0.0052s] [ 82%] 2024-08-06T21:15:51.1880758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 82%] 2024-08-06T21:15:51.1882089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 82%] 2024-08-06T21:15:51.1883385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0134s] [ 82%] 2024-08-06T21:15:51.1884660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0136s] [ 82%] 2024-08-06T21:15:51.1885995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 82%] 2024-08-06T21:15:51.1887328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 82%] 2024-08-06T21:15:51.1888598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0095s] [ 82%] 2024-08-06T21:15:51.1889897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0104s] [ 82%] 2024-08-06T21:15:51.1891194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 82%] 2024-08-06T21:15:51.1892554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 ('RERUN', {'yellow': True}) [0.0045s] [ 82%] 2024-08-06T21:15:51.1893885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 ('RERUN', {'yellow': True}) [0.0045s] [ 82%] 2024-08-06T21:15:51.1895183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 82%] 2024-08-06T21:15:51.1896464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0132s] [ 82%] 2024-08-06T21:15:51.1897757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0135s] [ 82%] 2024-08-06T21:15:51.1899029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 82%] 2024-08-06T21:15:51.1900371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 82%] 2024-08-06T21:15:51.1901740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0101s] [ 82%] 2024-08-06T21:15:51.1903081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 ('RERUN', {'yellow': True}) [0.0106s] [ 82%] 2024-08-06T21:15:51.1904883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 ('RERUN', {'yellow': True}) [0.0104s] [ 82%] 2024-08-06T21:15:51.1906262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 FAILED [0.0102s] [ 82%] 2024-08-06T21:15:51.1906288Z 2024-08-06T21:15:51.1906481Z ==================================== RERUNS ==================================== 2024-08-06T21:15:51.1907568Z _ TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 _ 2024-08-06T21:15:51.1907741Z Traceback (most recent call last): 2024-08-06T21:15:51.1908316Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 2930, in test_mem_efficient_attention_attn_mask_vs_math_ref_grads 2024-08-06T21:15:51.1908454Z check_out_and_grad( 2024-08-06T21:15:51.1908872Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 184, in check_out_and_grad 2024-08-06T21:15:51.1909254Z _check_equal(ref_grad, lp_ref_grad, comp_grad, fudge_factors.get(name, default_fudge_factor), name) 2024-08-06T21:15:51.1909619Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 144, in _check_equal 2024-08-06T21:15:51.1909773Z raise ValueError(msg) 2024-08-06T21:15:51.1910192Z ValueError: grad_query Test error 0.5072804925091745 is greater than threshold 1.4052779395874737e-05! 2024-08-06T21:15:51.1910211Z 2024-08-06T21:15:51.1910473Z To execute this test, run the following from the base repo dir: 2024-08-06T21:15:51.1911421Z python test/test_transformers.py -k TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 2024-08-06T21:15:51.1911440Z 2024-08-06T21:15:51.1911743Z This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0 2024-08-06T21:15:51.1912800Z _ TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 _ 2024-08-06T21:15:51.1912952Z Traceback (most recent call last): 2024-08-06T21:15:51.1913522Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 2930, in test_mem_efficient_attention_attn_mask_vs_math_ref_grads 2024-08-06T21:15:51.1913672Z check_out_and_grad( 2024-08-06T21:15:51.1914064Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 184, in check_out_and_grad 2024-08-06T21:15:51.1914457Z _check_equal(ref_grad, lp_ref_grad, comp_grad, fudge_factors.get(name, default_fudge_factor), name) 2024-08-06T21:15:51.1914820Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 144, in _check_equal 2024-08-06T21:15:51.1914957Z raise ValueError(msg) 2024-08-06T21:15:51.1915446Z ValueError: grad_query Test error 0.5072804925091745 is greater than threshold 1.4052779395874737e-05! 2024-08-06T21:15:51.1915512Z 2024-08-06T21:15:51.1915790Z To execute this test, run the following from the base repo dir: 2024-08-06T21:15:51.1916731Z python test/test_transformers.py -k TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 2024-08-06T21:15:51.1916765Z 2024-08-06T21:15:51.1917061Z This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0 2024-08-06T21:15:51.1918089Z _ TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 _ 2024-08-06T21:15:51.1918258Z Traceback (most recent call last): 2024-08-06T21:15:51.1918931Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 2930, in test_mem_efficient_attention_attn_mask_vs_math_ref_grads 2024-08-06T21:15:51.1919130Z check_out_and_grad( 2024-08-06T21:15:51.1919545Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 184, in check_out_and_grad 2024-08-06T21:15:51.1919924Z _check_equal(ref_grad, lp_ref_grad, comp_grad, fudge_factors.get(name, default_fudge_factor), name) 2024-08-06T21:15:51.1920303Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 144, in _check_equal 2024-08-06T21:15:51.1920439Z raise ValueError(msg) 2024-08-06T21:15:51.1920836Z ValueError: grad_query Test error 0.3257364332675934 is greater than threshold 0.0005705282092094421! 2024-08-06T21:15:51.1920854Z 2024-08-06T21:15:51.1921115Z To execute this test, run the following from the base repo dir: 2024-08-06T21:15:51.1922052Z python test/test_transformers.py -k TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 2024-08-06T21:15:51.1922078Z 2024-08-06T21:15:51.1922397Z This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0 2024-08-06T21:15:51.1923422Z _ TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 _ 2024-08-06T21:15:51.1923573Z Traceback (most recent call last): 2024-08-06T21:15:51.1924157Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 2930, in test_mem_efficient_attention_attn_mask_vs_math_ref_grads 2024-08-06T21:15:51.1924289Z check_out_and_grad( 2024-08-06T21:15:51.1924683Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 184, in check_out_and_grad 2024-08-06T21:15:51.1925074Z _check_equal(ref_grad, lp_ref_grad, comp_grad, fudge_factors.get(name, default_fudge_factor), name) 2024-08-06T21:15:51.1925440Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 144, in _check_equal 2024-08-06T21:15:51.1925582Z raise ValueError(msg) 2024-08-06T21:15:51.1926004Z ValueError: grad_query Test error 0.011955185793340206 is greater than threshold 0.000567873939871788! 2024-08-06T21:15:51.1926022Z 2024-08-06T21:15:51.1926262Z To execute this test, run the following from the base repo dir: 2024-08-06T21:15:51.1927222Z python test/test_transformers.py -k TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 2024-08-06T21:15:51.1927240Z 2024-08-06T21:15:51.1927531Z This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0 2024-08-06T21:15:51.1928553Z _ TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 _ 2024-08-06T21:15:51.1928768Z Traceback (most recent call last): 2024-08-06T21:15:51.1929385Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 2930, in test_mem_efficient_attention_attn_mask_vs_math_ref_grads 2024-08-06T21:15:51.1929537Z check_out_and_grad( 2024-08-06T21:15:51.1929930Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 184, in check_out_and_grad 2024-08-06T21:15:51.1930307Z _check_equal(ref_grad, lp_ref_grad, comp_grad, fudge_factors.get(name, default_fudge_factor), name) 2024-08-06T21:15:51.1930686Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 144, in _check_equal 2024-08-06T21:15:51.1930823Z raise ValueError(msg) 2024-08-06T21:15:51.1931183Z ValueError: grad_key Test error 7.938149022358619 is greater than threshold 2.792317079219092! 2024-08-06T21:15:51.1931201Z 2024-08-06T21:15:51.1931460Z To execute this test, run the following from the base repo dir: 2024-08-06T21:15:51.1932448Z python test/test_transformers.py -k TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 2024-08-06T21:15:51.1932505Z 2024-08-06T21:15:51.1932828Z This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0 2024-08-06T21:15:51.1933860Z _ TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 _ 2024-08-06T21:15:51.1934016Z Traceback (most recent call last): 2024-08-06T21:15:51.1934602Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 2930, in test_mem_efficient_attention_attn_mask_vs_math_ref_grads 2024-08-06T21:15:51.1934734Z check_out_and_grad( 2024-08-06T21:15:51.1935146Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 184, in check_out_and_grad 2024-08-06T21:15:51.1935531Z _check_equal(ref_grad, lp_ref_grad, comp_grad, fudge_factors.get(name, default_fudge_factor), name) 2024-08-06T21:15:51.1935899Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 144, in _check_equal 2024-08-06T21:15:51.1936058Z raise ValueError(msg) 2024-08-06T21:15:51.1936397Z ValueError: grad_query Test error 0.0023097957018762827 is greater than threshold 1e-05! 2024-08-06T21:15:51.1936415Z 2024-08-06T21:15:51.1936654Z To execute this test, run the following from the base repo dir: 2024-08-06T21:15:51.1937606Z python test/test_transformers.py -k TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 2024-08-06T21:15:51.1937624Z 2024-08-06T21:15:51.1937912Z This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0 2024-08-06T21:15:51.1938965Z _ TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 _ 2024-08-06T21:15:51.1939120Z Traceback (most recent call last): 2024-08-06T21:15:51.1939690Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 2930, in test_mem_efficient_attention_attn_mask_vs_math_ref_grads 2024-08-06T21:15:51.1939837Z check_out_and_grad( 2024-08-06T21:15:51.1940225Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 184, in check_out_and_grad 2024-08-06T21:15:51.1940600Z _check_equal(ref_grad, lp_ref_grad, comp_grad, fudge_factors.get(name, default_fudge_factor), name) 2024-08-06T21:15:51.1940977Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 135, in _check_equal 2024-08-06T21:15:51.1941141Z raise ValueError("Output/Grad with NaN") 2024-08-06T21:15:51.1941303Z ValueError: Output/Grad with NaN 2024-08-06T21:15:51.1941321Z 2024-08-06T21:15:51.1941562Z To execute this test, run the following from the base repo dir: 2024-08-06T21:15:51.1942554Z python test/test_transformers.py -k TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 2024-08-06T21:15:51.1942611Z 2024-08-06T21:15:51.1942932Z This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0 2024-08-06T21:15:51.1943964Z _ TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 _ 2024-08-06T21:15:51.1944128Z Traceback (most recent call last): 2024-08-06T21:15:51.1944694Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 2930, in test_mem_efficient_attention_attn_mask_vs_math_ref_grads 2024-08-06T21:15:51.1944825Z check_out_and_grad( 2024-08-06T21:15:51.1945231Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 184, in check_out_and_grad 2024-08-06T21:15:51.1945650Z _check_equal(ref_grad, lp_ref_grad, comp_grad, fudge_factors.get(name, default_fudge_factor), name) 2024-08-06T21:15:51.1946061Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 144, in _check_equal 2024-08-06T21:15:51.1946214Z raise ValueError(msg) 2024-08-06T21:15:51.1946572Z ValueError: grad_key Test error 23.15777015686035 is greater than threshold 2.763030767440796! 2024-08-06T21:15:51.1946590Z 2024-08-06T21:15:51.1946844Z To execute this test, run the following from the base repo dir: 2024-08-06T21:15:51.1947781Z python test/test_transformers.py -k TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 2024-08-06T21:15:51.1947800Z 2024-08-06T21:15:51.1948092Z This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0 2024-08-06T21:15:51.1949149Z _ TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 _ 2024-08-06T21:15:51.1949303Z Traceback (most recent call last): 2024-08-06T21:15:51.1949868Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 2930, in test_mem_efficient_attention_attn_mask_vs_math_ref_grads 2024-08-06T21:15:51.1950014Z check_out_and_grad( 2024-08-06T21:15:51.1950403Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 184, in check_out_and_grad 2024-08-06T21:15:51.1950802Z _check_equal(ref_grad, lp_ref_grad, comp_grad, fudge_factors.get(name, default_fudge_factor), name) 2024-08-06T21:15:51.1951163Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 144, in _check_equal 2024-08-06T21:15:51.1951297Z raise ValueError(msg) 2024-08-06T21:15:51.1951645Z ValueError: grad_query Test error 0.029102804139256477 is greater than threshold 1e-05! 2024-08-06T21:15:51.1951662Z 2024-08-06T21:15:51.1951904Z To execute this test, run the following from the base repo dir: 2024-08-06T21:15:51.1952847Z python test/test_transformers.py -k TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 2024-08-06T21:15:51.1952884Z 2024-08-06T21:15:51.1953178Z This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0 2024-08-06T21:15:51.1953353Z =================================== FAILURES =================================== 2024-08-06T21:15:51.1954404Z _ TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 _ 2024-08-06T21:15:51.1954554Z Traceback (most recent call last): 2024-08-06T21:15:51.1955120Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 2930, in test_mem_efficient_attention_attn_mask_vs_math_ref_grads 2024-08-06T21:15:51.1955312Z check_out_and_grad( 2024-08-06T21:15:51.1955755Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 184, in check_out_and_grad 2024-08-06T21:15:51.1956154Z _check_equal(ref_grad, lp_ref_grad, comp_grad, fudge_factors.get(name, default_fudge_factor), name) 2024-08-06T21:15:51.1956517Z File "/var/lib/jenkins/workspace/test/test_transformers.py", line 144, in _check_equal 2024-08-06T21:15:51.1956653Z raise ValueError(msg) 2024-08-06T21:15:51.1956996Z ValueError: grad_query Test error 22.65880012512207 is greater than threshold 1e-05! 2024-08-06T21:15:51.1957013Z 2024-08-06T21:15:51.1957253Z To execute this test, run the following from the base repo dir: 2024-08-06T21:15:51.1958193Z python test/test_transformers.py -k TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 2024-08-06T21:15:51.1958228Z 2024-08-06T21:15:51.1958572Z This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0 2024-08-06T21:15:51.1959291Z - generated xml file: /var/lib/jenkins/workspace/test/test-reports/python-pytest/test_transformers/test_transformers-cd06bb4e6af626fa.xml - 2024-08-06T21:15:51.1959665Z =========================== short test summary info ============================ 2024-08-06T21:15:51.1962003Z FAILED [0.0102s] test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 - ValueError: grad_query Test error 22.65880012512207 is greater than threshold 1e-05! 2024-08-06T21:15:51.1962026Z 2024-08-06T21:15:51.1962289Z To execute this test, run the following from the base repo dir: 2024-08-06T21:15:51.1963236Z python test/test_transformers.py -k TestSDPACudaOnlyCUDA.test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 2024-08-06T21:15:51.1963260Z 2024-08-06T21:15:51.1963555Z This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0 2024-08-06T21:15:51.1963928Z !!!!!!!!!!!!!!!!!!!!!!!!!! stopping after 1 failures !!!!!!!!!!!!!!!!!!!!!!!!!!! 2024-08-06T21:15:51.1964536Z ======= 1 failed, 208 passed, 37395 skipped, 9 rerun in 87.23s (0:01:27) ======= 2024-08-06T21:15:51.1964667Z Got exit code 1 2024-08-06T21:15:51.1964820Z Retrying single test... 2024-08-06T21:15:51.1965313Z Test results will be stored in test-reports/python-pytest/test_transformers/test_transformers-447fb59da2867f77.xml 2024-08-06T21:15:51.1965648Z ============================= test session starts ============================== 2024-08-06T21:15:51.1966020Z platform linux -- Python 3.10.14, pytest-7.3.2, pluggy-1.5.0 -- /opt/conda/envs/py_3.10/bin/python 2024-08-06T21:15:51.1966161Z cachedir: .pytest_cache 2024-08-06T21:15:51.1966721Z hypothesis profile 'pytorch_ci' -> database=None, max_examples=50, derandomize=True, suppress_health_check=[HealthCheck.too_slow] 2024-08-06T21:15:51.1966880Z rootdir: /var/lib/jenkins/workspace 2024-08-06T21:15:51.1967015Z configfile: pytest.ini 2024-08-06T21:15:51.1967461Z plugins: hypothesis-5.35.1, cpp-2.3.0, flakefinder-1.1.0, rerunfailures-14.0, xdist-3.3.1, xdoctest-1.1.0 2024-08-06T21:15:51.1967843Z collecting ... collected 45344 items / 45343 deselected / 1 selected 2024-08-06T21:15:51.1969021Z stepcurrent: skipping 37603 already run items. Running only test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 2024-08-06T21:15:51.1969183Z Running 1 items in this shard 2024-08-06T21:15:51.1969200Z 2024-08-06T21:15:51.1970566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.1291s] [100%] 2024-08-06T21:15:51.1970625Z 2024-08-06T21:15:51.1971318Z - generated xml file: /var/lib/jenkins/workspace/test/test-reports/python-pytest/test_transformers/test_transformers-447fb59da2867f77.xml - 2024-08-06T21:15:51.1971819Z ===================== 1 passed, 45343 deselected in 4.50s ====================== 2024-08-06T21:15:51.1971965Z Got exit code 0 2024-08-06T21:15:51.1972240Z Test succeeeded in new process, continuing with the rest of the tests 2024-08-06T21:15:51.1972729Z Test results will be stored in test-reports/python-pytest/test_transformers/test_transformers-00a63c3f2f793df3.xml 2024-08-06T21:15:51.1973062Z ============================= test session starts ============================== 2024-08-06T21:15:51.1973496Z platform linux -- Python 3.10.14, pytest-7.3.2, pluggy-1.5.0 -- /opt/conda/envs/py_3.10/bin/python 2024-08-06T21:15:51.1973679Z cachedir: .pytest_cache 2024-08-06T21:15:51.1974232Z hypothesis profile 'pytorch_ci' -> database=None, max_examples=50, derandomize=True, suppress_health_check=[HealthCheck.too_slow] 2024-08-06T21:15:51.1974383Z rootdir: /var/lib/jenkins/workspace 2024-08-06T21:15:51.1974520Z configfile: pytest.ini 2024-08-06T21:15:51.1974950Z plugins: hypothesis-5.35.1, cpp-2.3.0, flakefinder-1.1.0, rerunfailures-14.0, xdist-3.3.1, xdoctest-1.1.0 2024-08-06T21:15:51.1975354Z collecting ... collected 45344 items / 37604 deselected / 7740 selected 2024-08-06T21:15:51.1975558Z stepcurrent: skipping 37604 already run items. 2024-08-06T21:15:51.1975707Z Running 7740 items in this shard 2024-08-06T21:15:51.1975725Z 2024-08-06T21:15:51.1977020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.1297s] [ 0%] 2024-08-06T21:15:51.1978326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0093s] [ 0%] 2024-08-06T21:15:51.1979721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 [W806 21:13:37.009533287 attention.cpp:807] Warning: Dropout mask should only be used for testing purposes. (function operator()) 2024-08-06T21:15:51.1979954Z PASSED [0.1045s] [ 0%] 2024-08-06T21:15:51.1981242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0230s] [ 0%] 2024-08-06T21:15:51.1982525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0110s] [ 0%] 2024-08-06T21:15:51.1983826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0096s] [ 0%] 2024-08-06T21:15:51.1985094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0138s] [ 0%] 2024-08-06T21:15:51.1986439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0140s] [ 0%] 2024-08-06T21:15:51.1987756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0084s] [ 0%] 2024-08-06T21:15:51.1989040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 0%] 2024-08-06T21:15:51.1990313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0200s] [ 0%] 2024-08-06T21:15:51.1991646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0197s] [ 0%] 2024-08-06T21:15:51.1992970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0091s] [ 0%] 2024-08-06T21:15:51.1994261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0083s] [ 0%] 2024-08-06T21:15:51.1995539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0117s] [ 0%] 2024-08-06T21:15:51.1996815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 0%] 2024-08-06T21:15:51.1998096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0064s] [ 0%] 2024-08-06T21:15:51.1999366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 0%] 2024-08-06T21:15:51.2000654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0087s] [ 0%] 2024-08-06T21:15:51.2001983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0086s] [ 0%] 2024-08-06T21:15:51.2003272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0084s] [ 0%] 2024-08-06T21:15:51.2005033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 0%] 2024-08-06T21:15:51.2006443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0067s] [ 0%] 2024-08-06T21:15:51.2007725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 0%] 2024-08-06T21:15:51.2008989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0064s] [ 0%] 2024-08-06T21:15:51.2010341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 0%] 2024-08-06T21:15:51.2011671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0090s] [ 0%] 2024-08-06T21:15:51.2012962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0088s] [ 0%] 2024-08-06T21:15:51.2014226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 0%] 2024-08-06T21:15:51.2015519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 0%] 2024-08-06T21:15:51.2016801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0082s] [ 0%] 2024-08-06T21:15:51.2018086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0070s] [ 0%] 2024-08-06T21:15:51.2019348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 0%] 2024-08-06T21:15:51.2020643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 0%] 2024-08-06T21:15:51.2021908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0083s] [ 0%] 2024-08-06T21:15:51.2023185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0085s] [ 0%] 2024-08-06T21:15:51.2024522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 0%] 2024-08-06T21:15:51.2025853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 0%] 2024-08-06T21:15:51.2027133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 0%] 2024-08-06T21:15:51.2028402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 0%] 2024-08-06T21:15:51.2029725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 0%] 2024-08-06T21:15:51.2031039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 0%] 2024-08-06T21:15:51.2032317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 0%] 2024-08-06T21:15:51.2033595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 0%] 2024-08-06T21:15:51.2034866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 0%] 2024-08-06T21:15:51.2036175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 0%] 2024-08-06T21:15:51.2037443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 0%] 2024-08-06T21:15:51.2038730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 0%] 2024-08-06T21:15:51.2040000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 0%] 2024-08-06T21:15:51.2041282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 0%] 2024-08-06T21:15:51.2042589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 0%] 2024-08-06T21:15:51.2043943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 0%] 2024-08-06T21:15:51.2045209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 0%] 2024-08-06T21:15:51.2046519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 0%] 2024-08-06T21:15:51.2047826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 0%] 2024-08-06T21:15:51.2049144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 0%] 2024-08-06T21:15:51.2050423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 0%] 2024-08-06T21:15:51.2051693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 0%] 2024-08-06T21:15:51.2052980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0075s] [ 0%] 2024-08-06T21:15:51.2054254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 0%] 2024-08-06T21:15:51.2055535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 0%] 2024-08-06T21:15:51.2056829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 0%] 2024-08-06T21:15:51.2058115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 0%] 2024-08-06T21:15:51.2059380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 0%] 2024-08-06T21:15:51.2060643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 0%] 2024-08-06T21:15:51.2061973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 0%] 2024-08-06T21:15:51.2063289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 0%] 2024-08-06T21:15:51.2064578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0078s] [ 0%] 2024-08-06T21:15:51.2065860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 0%] 2024-08-06T21:15:51.2067196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 0%] 2024-08-06T21:15:51.2068503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 0%] 2024-08-06T21:15:51.2069782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 0%] 2024-08-06T21:15:51.2071041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 0%] 2024-08-06T21:15:51.2072327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 0%] 2024-08-06T21:15:51.2073590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0072s] [ 0%] 2024-08-06T21:15:51.2074855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 0%] 2024-08-06T21:15:51.2076161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 0%] 2024-08-06T21:15:51.2077435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 1%] 2024-08-06T21:15:51.2078727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0194s] [ 1%] 2024-08-06T21:15:51.2080048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0211s] [ 1%] 2024-08-06T21:15:51.2081386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0092s] [ 1%] 2024-08-06T21:15:51.2082668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0091s] [ 1%] 2024-08-06T21:15:51.2083958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0292s] [ 1%] 2024-08-06T21:15:51.2085274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0283s] [ 1%] 2024-08-06T21:15:51.2086615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0094s] [ 1%] 2024-08-06T21:15:51.2087912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0094s] [ 1%] 2024-08-06T21:15:51.2089183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0182s] [ 1%] 2024-08-06T21:15:51.2090484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0198s] [ 1%] 2024-08-06T21:15:51.2091753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0088s] [ 1%] 2024-08-06T21:15:51.2093045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0088s] [ 1%] 2024-08-06T21:15:51.2094323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0263s] [ 1%] 2024-08-06T21:15:51.2095645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0272s] [ 1%] 2024-08-06T21:15:51.2096915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0118s] [ 1%] 2024-08-06T21:15:51.2098206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0093s] [ 1%] 2024-08-06T21:15:51.2099518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0186s] [ 1%] 2024-08-06T21:15:51.2100843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0215s] [ 1%] 2024-08-06T21:15:51.2102164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0090s] [ 1%] 2024-08-06T21:15:51.2103449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0087s] [ 1%] 2024-08-06T21:15:51.2105188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0260s] [ 1%] 2024-08-06T21:15:51.2106587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0266s] [ 1%] 2024-08-06T21:15:51.2107877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0096s] [ 1%] 2024-08-06T21:15:51.2109158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0095s] [ 1%] 2024-08-06T21:15:51.2110449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0174s] [ 1%] 2024-08-06T21:15:51.2111716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0166s] [ 1%] 2024-08-06T21:15:51.2112980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0082s] [ 1%] 2024-08-06T21:15:51.2114274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 1%] 2024-08-06T21:15:51.2115546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0224s] [ 1%] 2024-08-06T21:15:51.2116855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0236s] [ 1%] 2024-08-06T21:15:51.2118195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0087s] [ 1%] 2024-08-06T21:15:51.2119558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0086s] [ 1%] 2024-08-06T21:15:51.2120820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 1%] 2024-08-06T21:15:51.2122105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 1%] 2024-08-06T21:15:51.2123413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 1%] 2024-08-06T21:15:51.2124756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 1%] 2024-08-06T21:15:51.2126021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0069s] [ 1%] 2024-08-06T21:15:51.2127296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 1%] 2024-08-06T21:15:51.2128585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 1%] 2024-08-06T21:15:51.2129858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 1%] 2024-08-06T21:15:51.2131136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 1%] 2024-08-06T21:15:51.2132408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 1%] 2024-08-06T21:15:51.2133698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 1%] 2024-08-06T21:15:51.2134959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 1%] 2024-08-06T21:15:51.2136262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0069s] [ 1%] 2024-08-06T21:15:51.2137580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 1%] 2024-08-06T21:15:51.2138913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 1%] 2024-08-06T21:15:51.2140181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 1%] 2024-08-06T21:15:51.2141490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 1%] 2024-08-06T21:15:51.2142830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 1%] 2024-08-06T21:15:51.2144091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 1%] 2024-08-06T21:15:51.2145372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 1%] 2024-08-06T21:15:51.2146666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0076s] [ 1%] 2024-08-06T21:15:51.2147963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 1%] 2024-08-06T21:15:51.2149232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 1%] 2024-08-06T21:15:51.2150523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 1%] 2024-08-06T21:15:51.2151796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 1%] 2024-08-06T21:15:51.2153063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 1%] 2024-08-06T21:15:51.2154340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 1%] 2024-08-06T21:15:51.2155654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 1%] 2024-08-06T21:15:51.2157003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0069s] [ 1%] 2024-08-06T21:15:51.2158277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 1%] 2024-08-06T21:15:51.2159554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 1%] 2024-08-06T21:15:51.2160864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 1%] 2024-08-06T21:15:51.2162199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 1%] 2024-08-06T21:15:51.2163467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 1%] 2024-08-06T21:15:51.2164753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 1%] 2024-08-06T21:15:51.2166034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 1%] 2024-08-06T21:15:51.2167299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0068s] [ 1%] 2024-08-06T21:15:51.2168581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 1%] 2024-08-06T21:15:51.2169843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 1%] 2024-08-06T21:15:51.2171131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 1%] 2024-08-06T21:15:51.2172393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 1%] 2024-08-06T21:15:51.2173716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 1%] 2024-08-06T21:15:51.2175029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 1%] 2024-08-06T21:15:51.2176343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 1%] 2024-08-06T21:15:51.2177607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0073s] [ 2%] 2024-08-06T21:15:51.2178922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0078s] [ 2%] 2024-08-06T21:15:51.2180266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 2%] 2024-08-06T21:15:51.2181534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 2%] 2024-08-06T21:15:51.2182815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 2%] 2024-08-06T21:15:51.2184088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 2%] 2024-08-06T21:15:51.2185379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 2%] 2024-08-06T21:15:51.2186664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 2%] 2024-08-06T21:15:51.2187946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0071s] [ 2%] 2024-08-06T21:15:51.2189218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 2%] 2024-08-06T21:15:51.2190482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 2%] 2024-08-06T21:15:51.2191764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 2%] 2024-08-06T21:15:51.2193069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 2%] 2024-08-06T21:15:51.2194451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 2%] 2024-08-06T21:15:51.2195706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 2%] 2024-08-06T21:15:51.2196994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 2%] 2024-08-06T21:15:51.2198296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 2%] 2024-08-06T21:15:51.2199642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 2%] 2024-08-06T21:15:51.2200905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 2%] 2024-08-06T21:15:51.2202232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 2%] 2024-08-06T21:15:51.2203511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0075s] [ 2%] 2024-08-06T21:15:51.2205169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 2%] 2024-08-06T21:15:51.2206493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 2%] 2024-08-06T21:15:51.2207826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 2%] 2024-08-06T21:15:51.2209129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0104s] [ 2%] 2024-08-06T21:15:51.2210407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0115s] [ 2%] 2024-08-06T21:15:51.2211772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 2%] 2024-08-06T21:15:51.2213121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 2%] 2024-08-06T21:15:51.2214414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0077s] [ 2%] 2024-08-06T21:15:51.2215688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0081s] [ 2%] 2024-08-06T21:15:51.2217061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 2%] 2024-08-06T21:15:51.2218408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 2%] 2024-08-06T21:15:51.2219679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0104s] [ 2%] 2024-08-06T21:15:51.2220973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 2%] 2024-08-06T21:15:51.2222247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 2%] 2024-08-06T21:15:51.2223553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 2%] 2024-08-06T21:15:51.2224826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0084s] [ 2%] 2024-08-06T21:15:51.2226121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0089s] [ 2%] 2024-08-06T21:15:51.2227402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 2%] 2024-08-06T21:15:51.2228701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 2%] 2024-08-06T21:15:51.2229970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0112s] [ 2%] 2024-08-06T21:15:51.2231292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0110s] [ 2%] 2024-08-06T21:15:51.2232629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 2%] 2024-08-06T21:15:51.2233903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 2%] 2024-08-06T21:15:51.2235191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0073s] [ 2%] 2024-08-06T21:15:51.2236523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 2%] 2024-08-06T21:15:51.2237862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 2%] 2024-08-06T21:15:51.2239130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 2%] 2024-08-06T21:15:51.2240424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0114s] [ 2%] 2024-08-06T21:15:51.2241709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 2%] 2024-08-06T21:15:51.2242998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 2%] 2024-08-06T21:15:51.2244273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 2%] 2024-08-06T21:15:51.2245547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0113s] [ 2%] 2024-08-06T21:15:51.2246868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0121s] [ 2%] 2024-08-06T21:15:51.2248135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0064s] [ 2%] 2024-08-06T21:15:51.2249468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 2%] 2024-08-06T21:15:51.2250790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0156s] [ 2%] 2024-08-06T21:15:51.2252088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0162s] [ 2%] 2024-08-06T21:15:51.2253355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0072s] [ 2%] 2024-08-06T21:15:51.2254729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 2%] 2024-08-06T21:15:51.2256066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0116s] [ 2%] 2024-08-06T21:15:51.2257360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0127s] [ 2%] 2024-08-06T21:15:51.2258650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0068s] [ 2%] 2024-08-06T21:15:51.2259931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 2%] 2024-08-06T21:15:51.2261222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0159s] [ 2%] 2024-08-06T21:15:51.2262498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0162s] [ 2%] 2024-08-06T21:15:51.2263809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0083s] [ 2%] 2024-08-06T21:15:51.2265097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0085s] [ 2%] 2024-08-06T21:15:51.2266397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0133s] [ 2%] 2024-08-06T21:15:51.2267670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0141s] [ 2%] 2024-08-06T21:15:51.2269005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0069s] [ 2%] 2024-08-06T21:15:51.2270337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 2%] 2024-08-06T21:15:51.2271609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0171s] [ 2%] 2024-08-06T21:15:51.2272909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0173s] [ 2%] 2024-08-06T21:15:51.2274227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0078s] [ 2%] 2024-08-06T21:15:51.2275575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0077s] [ 2%] 2024-08-06T21:15:51.2276863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0106s] [ 2%] 2024-08-06T21:15:51.2278162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0111s] [ 2%] 2024-08-06T21:15:51.2279434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0064s] [ 3%] 2024-08-06T21:15:51.2280724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 3%] 2024-08-06T21:15:51.2281999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0153s] [ 3%] 2024-08-06T21:15:51.2283286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0156s] [ 3%] 2024-08-06T21:15:51.2284576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0078s] [ 3%] 2024-08-06T21:15:51.2285852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0081s] [ 3%] 2024-08-06T21:15:51.2287201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0068s] [ 3%] 2024-08-06T21:15:51.2288536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 3%] 2024-08-06T21:15:51.2289816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 3%] 2024-08-06T21:15:51.2291081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 3%] 2024-08-06T21:15:51.2292409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0076s] [ 3%] 2024-08-06T21:15:51.2293738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 3%] 2024-08-06T21:15:51.2295025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 3%] 2024-08-06T21:15:51.2296312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 3%] 2024-08-06T21:15:51.2297585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 3%] 2024-08-06T21:15:51.2298876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 3%] 2024-08-06T21:15:51.2300138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 3%] 2024-08-06T21:15:51.2301429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 3%] 2024-08-06T21:15:51.2302739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 3%] 2024-08-06T21:15:51.2304381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 3%] 2024-08-06T21:15:51.2305686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 3%] 2024-08-06T21:15:51.2307079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 3%] 2024-08-06T21:15:51.2308424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 3%] 2024-08-06T21:15:51.2309689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 3%] 2024-08-06T21:15:51.2311023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 3%] 2024-08-06T21:15:51.2312357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 3%] 2024-08-06T21:15:51.2313645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0082s] [ 3%] 2024-08-06T21:15:51.2314914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 3%] 2024-08-06T21:15:51.2316201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 3%] 2024-08-06T21:15:51.2317497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 3%] 2024-08-06T21:15:51.2318778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0067s] [ 3%] 2024-08-06T21:15:51.2320044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 3%] 2024-08-06T21:15:51.2321330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 3%] 2024-08-06T21:15:51.2322598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 3%] 2024-08-06T21:15:51.2323859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 3%] 2024-08-06T21:15:51.2325191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 3%] 2024-08-06T21:15:51.2326526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 3%] 2024-08-06T21:15:51.2327815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 3%] 2024-08-06T21:15:51.2329077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 3%] 2024-08-06T21:15:51.2330403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 3%] 2024-08-06T21:15:51.2331713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 3%] 2024-08-06T21:15:51.2332991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 3%] 2024-08-06T21:15:51.2334254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0072s] [ 3%] 2024-08-06T21:15:51.2335527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 3%] 2024-08-06T21:15:51.2336826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 3%] 2024-08-06T21:15:51.2338095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 3%] 2024-08-06T21:15:51.2339381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 3%] 2024-08-06T21:15:51.2340656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 3%] 2024-08-06T21:15:51.2341933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 3%] 2024-08-06T21:15:51.2343197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 3%] 2024-08-06T21:15:51.2344518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0075s] [ 3%] 2024-08-06T21:15:51.2345836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 3%] 2024-08-06T21:15:51.2347139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 3%] 2024-08-06T21:15:51.2348451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 3%] 2024-08-06T21:15:51.2349768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 3%] 2024-08-06T21:15:51.2351053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 3%] 2024-08-06T21:15:51.2352311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 3%] 2024-08-06T21:15:51.2353595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 3%] 2024-08-06T21:15:51.2354864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0080s] [ 3%] 2024-08-06T21:15:51.2356146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 3%] 2024-08-06T21:15:51.2357429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 3%] 2024-08-06T21:15:51.2358722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 3%] 2024-08-06T21:15:51.2359983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 3%] 2024-08-06T21:15:51.2361244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 3%] 2024-08-06T21:15:51.2362567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 3%] 2024-08-06T21:15:51.2363885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 3%] 2024-08-06T21:15:51.2365160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0071s] [ 3%] 2024-08-06T21:15:51.2366441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 3%] 2024-08-06T21:15:51.2367763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 3%] 2024-08-06T21:15:51.2369081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 3%] 2024-08-06T21:15:51.2370366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0194s] [ 3%] 2024-08-06T21:15:51.2371639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0211s] [ 3%] 2024-08-06T21:15:51.2372932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0086s] [ 3%] 2024-08-06T21:15:51.2374208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 3%] 2024-08-06T21:15:51.2375476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0280s] [ 3%] 2024-08-06T21:15:51.2376802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0296s] [ 3%] 2024-08-06T21:15:51.2378083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0100s] [ 3%] 2024-08-06T21:15:51.2379379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0096s] [ 4%] 2024-08-06T21:15:51.2380685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0201s] [ 4%] 2024-08-06T21:15:51.2382028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0227s] [ 4%] 2024-08-06T21:15:51.2383298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0091s] [ 4%] 2024-08-06T21:15:51.2384587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0092s] [ 4%] 2024-08-06T21:15:51.2385903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0291s] [ 4%] 2024-08-06T21:15:51.2387255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0300s] [ 4%] 2024-08-06T21:15:51.2388543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0109s] [ 4%] 2024-08-06T21:15:51.2389816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0108s] [ 4%] 2024-08-06T21:15:51.2391113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0235s] [ 4%] 2024-08-06T21:15:51.2392395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0257s] [ 4%] 2024-08-06T21:15:51.2393679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0105s] [ 4%] 2024-08-06T21:15:51.2394957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0106s] [ 4%] 2024-08-06T21:15:51.2396251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0312s] [ 4%] 2024-08-06T21:15:51.2397533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0320s] [ 4%] 2024-08-06T21:15:51.2398818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0116s] [ 4%] 2024-08-06T21:15:51.2400155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0108s] [ 4%] 2024-08-06T21:15:51.2401479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0185s] [ 4%] 2024-08-06T21:15:51.2402805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0201s] [ 4%] 2024-08-06T21:15:51.2404436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0087s] [ 4%] 2024-08-06T21:15:51.2405842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 4%] 2024-08-06T21:15:51.2407203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0274s] [ 4%] 2024-08-06T21:15:51.2408501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0285s] [ 4%] 2024-08-06T21:15:51.2409775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0095s] [ 4%] 2024-08-06T21:15:51.2411079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0093s] [ 4%] 2024-08-06T21:15:51.2412345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 4%] 2024-08-06T21:15:51.2413611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 4%] 2024-08-06T21:15:51.2414897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 4%] 2024-08-06T21:15:51.2416176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 4%] 2024-08-06T21:15:51.2417488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0076s] [ 4%] 2024-08-06T21:15:51.2418826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 4%] 2024-08-06T21:15:51.2420173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 4%] 2024-08-06T21:15:51.2421454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 4%] 2024-08-06T21:15:51.2422735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 4%] 2024-08-06T21:15:51.2424046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 4%] 2024-08-06T21:15:51.2425380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 4%] 2024-08-06T21:15:51.2426645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 4%] 2024-08-06T21:15:51.2427917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0077s] [ 4%] 2024-08-06T21:15:51.2429217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0078s] [ 4%] 2024-08-06T21:15:51.2430492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 4%] 2024-08-06T21:15:51.2431780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 4%] 2024-08-06T21:15:51.2433046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0062s] [ 4%] 2024-08-06T21:15:51.2434340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 4%] 2024-08-06T21:15:51.2435600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 4%] 2024-08-06T21:15:51.2436900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 4%] 2024-08-06T21:15:51.2438216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0080s] [ 4%] 2024-08-06T21:15:51.2439563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 4%] 2024-08-06T21:15:51.2440825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 4%] 2024-08-06T21:15:51.2442102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 4%] 2024-08-06T21:15:51.2443424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 4%] 2024-08-06T21:15:51.2444741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 4%] 2024-08-06T21:15:51.2446024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 4%] 2024-08-06T21:15:51.2447324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 4%] 2024-08-06T21:15:51.2448617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 4%] 2024-08-06T21:15:51.2449886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 4%] 2024-08-06T21:15:51.2451164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 4%] 2024-08-06T21:15:51.2452439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 4%] 2024-08-06T21:15:51.2453706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 4%] 2024-08-06T21:15:51.2454989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 4%] 2024-08-06T21:15:51.2456293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 4%] 2024-08-06T21:15:51.2457649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 4%] 2024-08-06T21:15:51.2458918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 4%] 2024-08-06T21:15:51.2460205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 4%] 2024-08-06T21:15:51.2461510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 4%] 2024-08-06T21:15:51.2462868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 4%] 2024-08-06T21:15:51.2464130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 4%] 2024-08-06T21:15:51.2465415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 4%] 2024-08-06T21:15:51.2466698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 4%] 2024-08-06T21:15:51.2467977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 4%] 2024-08-06T21:15:51.2469256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0075s] [ 4%] 2024-08-06T21:15:51.2470526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 4%] 2024-08-06T21:15:51.2471815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 4%] 2024-08-06T21:15:51.2473087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 4%] 2024-08-06T21:15:51.2474359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 4%] 2024-08-06T21:15:51.2475672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 4%] 2024-08-06T21:15:51.2477022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 4%] 2024-08-06T21:15:51.2478292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 4%] 2024-08-06T21:15:51.2479561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0079s] [ 5%] 2024-08-06T21:15:51.2480891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0078s] [ 5%] 2024-08-06T21:15:51.2482203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 5%] 2024-08-06T21:15:51.2483488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 5%] 2024-08-06T21:15:51.2484750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 5%] 2024-08-06T21:15:51.2486037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 5%] 2024-08-06T21:15:51.2487316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 5%] 2024-08-06T21:15:51.2488595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 5%] 2024-08-06T21:15:51.2489864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0071s] [ 5%] 2024-08-06T21:15:51.2491142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 5%] 2024-08-06T21:15:51.2492426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 5%] 2024-08-06T21:15:51.2493788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 5%] 2024-08-06T21:15:51.2495130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0089s] [ 5%] 2024-08-06T21:15:51.2496415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0094s] [ 5%] 2024-08-06T21:15:51.2497719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 5%] 2024-08-06T21:15:51.2499036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 5%] 2024-08-06T21:15:51.2500383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0130s] [ 5%] 2024-08-06T21:15:51.2501662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0139s] [ 5%] 2024-08-06T21:15:51.2502993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0073s] [ 5%] 2024-08-06T21:15:51.2504629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 5%] 2024-08-06T21:15:51.2505934Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0091s] [ 5%] 2024-08-06T21:15:51.2507276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0098s] [ 5%] 2024-08-06T21:15:51.2508561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0065s] [ 5%] 2024-08-06T21:15:51.2509860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 5%] 2024-08-06T21:15:51.2511131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0132s] [ 5%] 2024-08-06T21:15:51.2512426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0133s] [ 5%] 2024-08-06T21:15:51.2513779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 5%] 2024-08-06T21:15:51.2515148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 5%] 2024-08-06T21:15:51.2516419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0096s] [ 5%] 2024-08-06T21:15:51.2517741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 5%] 2024-08-06T21:15:51.2519069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 5%] 2024-08-06T21:15:51.2520409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 5%] 2024-08-06T21:15:51.2521698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0136s] [ 5%] 2024-08-06T21:15:51.2522975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0137s] [ 5%] 2024-08-06T21:15:51.2524270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 5%] 2024-08-06T21:15:51.2525549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 5%] 2024-08-06T21:15:51.2526835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0086s] [ 5%] 2024-08-06T21:15:51.2528128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0089s] [ 5%] 2024-08-06T21:15:51.2529405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 5%] 2024-08-06T21:15:51.2530672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 5%] 2024-08-06T21:15:51.2531985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0128s] [ 5%] 2024-08-06T21:15:51.2533328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0129s] [ 5%] 2024-08-06T21:15:51.2534594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 5%] 2024-08-06T21:15:51.2535882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 5%] 2024-08-06T21:15:51.2537214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0139s] [ 5%] 2024-08-06T21:15:51.2538562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0152s] [ 5%] 2024-08-06T21:15:51.2539828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0072s] [ 5%] 2024-08-06T21:15:51.2541119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 5%] 2024-08-06T21:15:51.2542395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0212s] [ 5%] 2024-08-06T21:15:51.2543690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0219s] [ 5%] 2024-08-06T21:15:51.2544959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0080s] [ 5%] 2024-08-06T21:15:51.2546239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 5%] 2024-08-06T21:15:51.2547553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0143s] [ 5%] 2024-08-06T21:15:51.2548829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0160s] [ 5%] 2024-08-06T21:15:51.2558770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0074s] [ 5%] 2024-08-06T21:15:51.2560329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 5%] 2024-08-06T21:15:51.2561700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0214s] [ 5%] 2024-08-06T21:15:51.2562972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0219s] [ 5%] 2024-08-06T21:15:51.2564288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0082s] [ 5%] 2024-08-06T21:15:51.2566214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 5%] 2024-08-06T21:15:51.2567494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0157s] [ 5%] 2024-08-06T21:15:51.2568756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0182s] [ 5%] 2024-08-06T21:15:51.2570025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0086s] [ 5%] 2024-08-06T21:15:51.2571298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0086s] [ 5%] 2024-08-06T21:15:51.2572559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0233s] [ 5%] 2024-08-06T21:15:51.2573832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0234s] [ 5%] 2024-08-06T21:15:51.2575099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0094s] [ 5%] 2024-08-06T21:15:51.2576388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0090s] [ 5%] 2024-08-06T21:15:51.2577642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0137s] [ 5%] 2024-08-06T21:15:51.2578951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0145s] [ 5%] 2024-08-06T21:15:51.2580259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0073s] [ 5%] 2024-08-06T21:15:51.2581524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 5%] 2024-08-06T21:15:51.2582779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0210s] [ 5%] 2024-08-06T21:15:51.2584087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0215s] [ 5%] 2024-08-06T21:15:51.2585403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0079s] [ 5%] 2024-08-06T21:15:51.2586692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0086s] [ 5%] 2024-08-06T21:15:51.2587959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 5%] 2024-08-06T21:15:51.2589224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 5%] 2024-08-06T21:15:51.2590482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 6%] 2024-08-06T21:15:51.2591735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 6%] 2024-08-06T21:15:51.2593000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0093s] [ 6%] 2024-08-06T21:15:51.2594266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0092s] [ 6%] 2024-08-06T21:15:51.2595527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 6%] 2024-08-06T21:15:51.2596850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 6%] 2024-08-06T21:15:51.2598160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0066s] [ 6%] 2024-08-06T21:15:51.2599426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 6%] 2024-08-06T21:15:51.2600676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 6%] 2024-08-06T21:15:51.2602047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 6%] 2024-08-06T21:15:51.2603370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0093s] [ 6%] 2024-08-06T21:15:51.2605104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0095s] [ 6%] 2024-08-06T21:15:51.2606408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 6%] 2024-08-06T21:15:51.2607684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 6%] 2024-08-06T21:15:51.2608944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0071s] [ 6%] 2024-08-06T21:15:51.2610202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 6%] 2024-08-06T21:15:51.2611465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0065s] [ 6%] 2024-08-06T21:15:51.2612728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 6%] 2024-08-06T21:15:51.2613989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0100s] [ 6%] 2024-08-06T21:15:51.2615252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0102s] [ 6%] 2024-08-06T21:15:51.2616638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 6%] 2024-08-06T21:15:51.2618007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 6%] 2024-08-06T21:15:51.2619271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 6%] 2024-08-06T21:15:51.2620528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 6%] 2024-08-06T21:15:51.2621855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 6%] 2024-08-06T21:15:51.2623179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 6%] 2024-08-06T21:15:51.2624435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0090s] [ 6%] 2024-08-06T21:15:51.2625709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0090s] [ 6%] 2024-08-06T21:15:51.2626997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 6%] 2024-08-06T21:15:51.2628262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 6%] 2024-08-06T21:15:51.2629511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0062s] [ 6%] 2024-08-06T21:15:51.2630783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 6%] 2024-08-06T21:15:51.2632045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 6%] 2024-08-06T21:15:51.2633307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 6%] 2024-08-06T21:15:51.2634604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0085s] [ 6%] 2024-08-06T21:15:51.2635916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 6%] 2024-08-06T21:15:51.2637191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 6%] 2024-08-06T21:15:51.2638447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 6%] 2024-08-06T21:15:51.2639745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0064s] [ 6%] 2024-08-06T21:15:51.2641058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 6%] 2024-08-06T21:15:51.2642313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 6%] 2024-08-06T21:15:51.2643565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 6%] 2024-08-06T21:15:51.2644833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0086s] [ 6%] 2024-08-06T21:15:51.2646110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0086s] [ 6%] 2024-08-06T21:15:51.2647372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 6%] 2024-08-06T21:15:51.2648634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 6%] 2024-08-06T21:15:51.2649895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0069s] [ 6%] 2024-08-06T21:15:51.2651153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 6%] 2024-08-06T21:15:51.2652400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 6%] 2024-08-06T21:15:51.2653703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 6%] 2024-08-06T21:15:51.2655009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0091s] [ 6%] 2024-08-06T21:15:51.2656299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0090s] [ 6%] 2024-08-06T21:15:51.2657559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 6%] 2024-08-06T21:15:51.2658868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 6%] 2024-08-06T21:15:51.2660168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 6%] 2024-08-06T21:15:51.2661423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 6%] 2024-08-06T21:15:51.2662686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 6%] 2024-08-06T21:15:51.2663947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 6%] 2024-08-06T21:15:51.2665204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0083s] [ 6%] 2024-08-06T21:15:51.2666480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0083s] [ 6%] 2024-08-06T21:15:51.2667746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 6%] 2024-08-06T21:15:51.2669006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 6%] 2024-08-06T21:15:51.2670271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0259s] [ 6%] 2024-08-06T21:15:51.2671579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0286s] [ 6%] 2024-08-06T21:15:51.2672900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0102s] [ 6%] 2024-08-06T21:15:51.2674168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0100s] [ 6%] 2024-08-06T21:15:51.2675430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0392s] [ 6%] 2024-08-06T21:15:51.2676751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0410s] [ 6%] 2024-08-06T21:15:51.2678070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0110s] [ 6%] 2024-08-06T21:15:51.2679346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0110s] [ 6%] 2024-08-06T21:15:51.2680602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0261s] [ 6%] 2024-08-06T21:15:51.2681878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0295s] [ 6%] 2024-08-06T21:15:51.2683142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0106s] [ 6%] 2024-08-06T21:15:51.2684414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0104s] [ 6%] 2024-08-06T21:15:51.2685695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0397s] [ 6%] 2024-08-06T21:15:51.2686971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0414s] [ 6%] 2024-08-06T21:15:51.2688239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0116s] [ 6%] 2024-08-06T21:15:51.2689503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0115s] [ 7%] 2024-08-06T21:15:51.2690830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0288s] [ 7%] 2024-08-06T21:15:51.2692148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0326s] [ 7%] 2024-08-06T21:15:51.2693412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0114s] [ 7%] 2024-08-06T21:15:51.2694677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0112s] [ 7%] 2024-08-06T21:15:51.2696011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0418s] [ 7%] 2024-08-06T21:15:51.2697327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0431s] [ 7%] 2024-08-06T21:15:51.2698594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0125s] [ 7%] 2024-08-06T21:15:51.2699862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0124s] [ 7%] 2024-08-06T21:15:51.2701130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0244s] [ 7%] 2024-08-06T21:15:51.2702438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0264s] [ 7%] 2024-08-06T21:15:51.2703701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0095s] [ 7%] 2024-08-06T21:15:51.2705380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0096s] [ 7%] 2024-08-06T21:15:51.2706686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0388s] [ 7%] 2024-08-06T21:15:51.2707957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0403s] [ 7%] 2024-08-06T21:15:51.2709299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0108s] [ 7%] 2024-08-06T21:15:51.2710648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0109s] [ 7%] 2024-08-06T21:15:51.2711900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 7%] 2024-08-06T21:15:51.2713155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 7%] 2024-08-06T21:15:51.2714468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 7%] 2024-08-06T21:15:51.2715816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 7%] 2024-08-06T21:15:51.2717078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0089s] [ 7%] 2024-08-06T21:15:51.2718343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0090s] [ 7%] 2024-08-06T21:15:51.2719610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 7%] 2024-08-06T21:15:51.2720874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 7%] 2024-08-06T21:15:51.2722128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0066s] [ 7%] 2024-08-06T21:15:51.2723388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 7%] 2024-08-06T21:15:51.2724655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0067s] [ 7%] 2024-08-06T21:15:51.2725936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 7%] 2024-08-06T21:15:51.2727189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0097s] [ 7%] 2024-08-06T21:15:51.2728509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0094s] [ 7%] 2024-08-06T21:15:51.2729822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0075s] [ 7%] 2024-08-06T21:15:51.2731086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 7%] 2024-08-06T21:15:51.2732383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0072s] [ 7%] 2024-08-06T21:15:51.2733695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 7%] 2024-08-06T21:15:51.2734945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 7%] 2024-08-06T21:15:51.2736219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 7%] 2024-08-06T21:15:51.2737479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0096s] [ 7%] 2024-08-06T21:15:51.2738753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0095s] [ 7%] 2024-08-06T21:15:51.2740009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 7%] 2024-08-06T21:15:51.2741276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 7%] 2024-08-06T21:15:51.2742537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0064s] [ 7%] 2024-08-06T21:15:51.2743797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 7%] 2024-08-06T21:15:51.2745062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 7%] 2024-08-06T21:15:51.2746364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 7%] 2024-08-06T21:15:51.2747679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0090s] [ 7%] 2024-08-06T21:15:51.2748937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0088s] [ 7%] 2024-08-06T21:15:51.2750194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 7%] 2024-08-06T21:15:51.2751495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 7%] 2024-08-06T21:15:51.2752815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 7%] 2024-08-06T21:15:51.2754076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 7%] 2024-08-06T21:15:51.2755345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 7%] 2024-08-06T21:15:51.2756613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 7%] 2024-08-06T21:15:51.2757870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0087s] [ 7%] 2024-08-06T21:15:51.2759133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0085s] [ 7%] 2024-08-06T21:15:51.2760388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 7%] 2024-08-06T21:15:51.2761656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 7%] 2024-08-06T21:15:51.2762905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 7%] 2024-08-06T21:15:51.2764157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 7%] 2024-08-06T21:15:51.2765480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 7%] 2024-08-06T21:15:51.2766792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 7%] 2024-08-06T21:15:51.2768050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0090s] [ 7%] 2024-08-06T21:15:51.2769352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0091s] [ 7%] 2024-08-06T21:15:51.2770659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 7%] 2024-08-06T21:15:51.2771913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 7%] 2024-08-06T21:15:51.2773166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0072s] [ 7%] 2024-08-06T21:15:51.2774422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 7%] 2024-08-06T21:15:51.2775695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 7%] 2024-08-06T21:15:51.2776947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 7%] 2024-08-06T21:15:51.2778200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0096s] [ 7%] 2024-08-06T21:15:51.2779474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0098s] [ 7%] 2024-08-06T21:15:51.2780730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 7%] 2024-08-06T21:15:51.2781992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 7%] 2024-08-06T21:15:51.2783278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0062s] [ 7%] 2024-08-06T21:15:51.2784605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 7%] 2024-08-06T21:15:51.2785853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 7%] 2024-08-06T21:15:51.2787108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 7%] 2024-08-06T21:15:51.2788406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0088s] [ 7%] 2024-08-06T21:15:51.2789768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0087s] [ 8%] 2024-08-06T21:15:51.2791022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 8%] 2024-08-06T21:15:51.2792276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 8%] 2024-08-06T21:15:51.2793534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0094s] [ 8%] 2024-08-06T21:15:51.2794815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 8%] 2024-08-06T21:15:51.2796066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0095s] [ 8%] 2024-08-06T21:15:51.2797324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 8%] 2024-08-06T21:15:51.2798589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 8%] 2024-08-06T21:15:51.2799850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 8%] 2024-08-06T21:15:51.2801106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 8%] 2024-08-06T21:15:51.2802450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 8%] 2024-08-06T21:15:51.2803764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 8%] 2024-08-06T21:15:51.2805449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 8%] 2024-08-06T21:15:51.2806830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 8%] 2024-08-06T21:15:51.2808175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 8%] 2024-08-06T21:15:51.2809427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 8%] 2024-08-06T21:15:51.2810692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 8%] 2024-08-06T21:15:51.2811951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 8%] 2024-08-06T21:15:51.2813224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 8%] 2024-08-06T21:15:51.2814480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 8%] 2024-08-06T21:15:51.2815742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 8%] 2024-08-06T21:15:51.2817010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 8%] 2024-08-06T21:15:51.2818268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 8%] 2024-08-06T21:15:51.2819522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 8%] 2024-08-06T21:15:51.2820845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 8%] 2024-08-06T21:15:51.2822174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 8%] 2024-08-06T21:15:51.2823433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 8%] 2024-08-06T21:15:51.2824705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 8%] 2024-08-06T21:15:51.2826023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 8%] 2024-08-06T21:15:51.2827331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 8%] 2024-08-06T21:15:51.2828586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 8%] 2024-08-06T21:15:51.2829838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 8%] 2024-08-06T21:15:51.2831109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 8%] 2024-08-06T21:15:51.2832362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 8%] 2024-08-06T21:15:51.2833621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 8%] 2024-08-06T21:15:51.2834896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 8%] 2024-08-06T21:15:51.2836166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 8%] 2024-08-06T21:15:51.2837416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 8%] 2024-08-06T21:15:51.2838679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 8%] 2024-08-06T21:15:51.2839978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0075s] [ 8%] 2024-08-06T21:15:51.2841295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 8%] 2024-08-06T21:15:51.2842551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 8%] 2024-08-06T21:15:51.2843851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 8%] 2024-08-06T21:15:51.2845172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0070s] [ 8%] 2024-08-06T21:15:51.2846429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 8%] 2024-08-06T21:15:51.2847684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 8%] 2024-08-06T21:15:51.2848941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 8%] 2024-08-06T21:15:51.2850207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 8%] 2024-08-06T21:15:51.2851460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 8%] 2024-08-06T21:15:51.2852717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 8%] 2024-08-06T21:15:51.2853989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 8%] 2024-08-06T21:15:51.2855243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0078s] [ 8%] 2024-08-06T21:15:51.2856503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 8%] 2024-08-06T21:15:51.2857796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0064s] [ 8%] 2024-08-06T21:15:51.2859111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 8%] 2024-08-06T21:15:51.2860363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0083s] [ 8%] 2024-08-06T21:15:51.2861624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 8%] 2024-08-06T21:15:51.2862915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 8%] 2024-08-06T21:15:51.2864259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 8%] 2024-08-06T21:15:51.2865513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 8%] 2024-08-06T21:15:51.2866768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 8%] 2024-08-06T21:15:51.2868025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 8%] 2024-08-06T21:15:51.2869279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 8%] 2024-08-06T21:15:51.2870537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0069s] [ 8%] 2024-08-06T21:15:51.2871799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 8%] 2024-08-06T21:15:51.2873060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 8%] 2024-08-06T21:15:51.2874337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 8%] 2024-08-06T21:15:51.2875585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 8%] 2024-08-06T21:15:51.2876882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 8%] 2024-08-06T21:15:51.2878190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 8%] 2024-08-06T21:15:51.2879440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 8%] 2024-08-06T21:15:51.2880692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 8%] 2024-08-06T21:15:51.2881999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 8%] 2024-08-06T21:15:51.2883298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 8%] 2024-08-06T21:15:51.2884575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 8%] 2024-08-06T21:15:51.2885822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 8%] 2024-08-06T21:15:51.2887081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 8%] 2024-08-06T21:15:51.2888322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 9%] 2024-08-06T21:15:51.2889575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 9%] 2024-08-06T21:15:51.2890827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 9%] 2024-08-06T21:15:51.2892087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 9%] 2024-08-06T21:15:51.2893347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 9%] 2024-08-06T21:15:51.2894642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 9%] 2024-08-06T21:15:51.2895964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 9%] 2024-08-06T21:15:51.2897212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 9%] 2024-08-06T21:15:51.2898462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 9%] 2024-08-06T21:15:51.2899755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 9%] 2024-08-06T21:15:51.2901061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 9%] 2024-08-06T21:15:51.2902353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 9%] 2024-08-06T21:15:51.2903637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 9%] 2024-08-06T21:15:51.2905279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 9%] 2024-08-06T21:15:51.2906544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 9%] 2024-08-06T21:15:51.2907795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 9%] 2024-08-06T21:15:51.2909037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 9%] 2024-08-06T21:15:51.2910295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 9%] 2024-08-06T21:15:51.2911537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 9%] 2024-08-06T21:15:51.2912790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 9%] 2024-08-06T21:15:51.2914135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 9%] 2024-08-06T21:15:51.2915460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 9%] 2024-08-06T21:15:51.2916698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 9%] 2024-08-06T21:15:51.2917944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 9%] 2024-08-06T21:15:51.2919249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 9%] 2024-08-06T21:15:51.2920559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 9%] 2024-08-06T21:15:51.2921810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 9%] 2024-08-06T21:15:51.2923062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 9%] 2024-08-06T21:15:51.2924341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 9%] 2024-08-06T21:15:51.2925590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 9%] 2024-08-06T21:15:51.2926834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 9%] 2024-08-06T21:15:51.2928081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 9%] 2024-08-06T21:15:51.2929327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 9%] 2024-08-06T21:15:51.2930576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 9%] 2024-08-06T21:15:51.2931856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 9%] 2024-08-06T21:15:51.2933164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 9%] 2024-08-06T21:15:51.2934422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 9%] 2024-08-06T21:15:51.2935673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 9%] 2024-08-06T21:15:51.2936956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 9%] 2024-08-06T21:15:51.2938259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 9%] 2024-08-06T21:15:51.2939496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 9%] 2024-08-06T21:15:51.2940744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 9%] 2024-08-06T21:15:51.2941994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 9%] 2024-08-06T21:15:51.2943259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 9%] 2024-08-06T21:15:51.2944511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 9%] 2024-08-06T21:15:51.2945760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 9%] 2024-08-06T21:15:51.2947013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 9%] 2024-08-06T21:15:51.2948262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 9%] 2024-08-06T21:15:51.2949498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 9%] 2024-08-06T21:15:51.2950785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 9%] 2024-08-06T21:15:51.2952078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 9%] 2024-08-06T21:15:51.2953344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 9%] 2024-08-06T21:15:51.2954580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 9%] 2024-08-06T21:15:51.2955875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 9%] 2024-08-06T21:15:51.2957180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0083s] [ 9%] 2024-08-06T21:15:51.2958438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0086s] [ 9%] 2024-08-06T21:15:51.2959685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 9%] 2024-08-06T21:15:51.2960948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 9%] 2024-08-06T21:15:51.2962200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0095s] [ 9%] 2024-08-06T21:15:51.2963489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0094s] [ 9%] 2024-08-06T21:15:51.2964742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 9%] 2024-08-06T21:15:51.2966005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 9%] 2024-08-06T21:15:51.2967265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0089s] [ 9%] 2024-08-06T21:15:51.2968515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0093s] [ 9%] 2024-08-06T21:15:51.2969827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 9%] 2024-08-06T21:15:51.2971140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 9%] 2024-08-06T21:15:51.2972399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0102s] [ 9%] 2024-08-06T21:15:51.2973713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0101s] [ 9%] 2024-08-06T21:15:51.2975024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 9%] 2024-08-06T21:15:51.2976278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 9%] 2024-08-06T21:15:51.2977533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0105s] [ 9%] 2024-08-06T21:15:51.2978791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0108s] [ 9%] 2024-08-06T21:15:51.2980049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 9%] 2024-08-06T21:15:51.2981307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 9%] 2024-08-06T21:15:51.2982570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0117s] [ 9%] 2024-08-06T21:15:51.2983847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0120s] [ 9%] 2024-08-06T21:15:51.2985105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 9%] 2024-08-06T21:15:51.2986367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 10%] 2024-08-06T21:15:51.2987654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0081s] [ 10%] 2024-08-06T21:15:51.2988961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 10%] 2024-08-06T21:15:51.2990204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 10%] 2024-08-06T21:15:51.2991452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 10%] 2024-08-06T21:15:51.2992777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0091s] [ 10%] 2024-08-06T21:15:51.2994085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0093s] [ 10%] 2024-08-06T21:15:51.2995338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 10%] 2024-08-06T21:15:51.2996589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 10%] 2024-08-06T21:15:51.2997844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 10%] 2024-08-06T21:15:51.2999095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 10%] 2024-08-06T21:15:51.3000341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 10%] 2024-08-06T21:15:51.3001594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 10%] 2024-08-06T21:15:51.3002919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 10%] 2024-08-06T21:15:51.3004553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 10%] 2024-08-06T21:15:51.3005833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 10%] 2024-08-06T21:15:51.3007172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 10%] 2024-08-06T21:15:51.3008485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 10%] 2024-08-06T21:15:51.3009743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 10%] 2024-08-06T21:15:51.3011060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 10%] 2024-08-06T21:15:51.3012406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 10%] 2024-08-06T21:15:51.3013679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 10%] 2024-08-06T21:15:51.3014956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 10%] 2024-08-06T21:15:51.3016217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 10%] 2024-08-06T21:15:51.3017485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 10%] 2024-08-06T21:15:51.3018758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 10%] 2024-08-06T21:15:51.3020017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 10%] 2024-08-06T21:15:51.3021294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 10%] 2024-08-06T21:15:51.3022570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 10%] 2024-08-06T21:15:51.3023847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 10%] 2024-08-06T21:15:51.3025155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 10%] 2024-08-06T21:15:51.3026483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 10%] 2024-08-06T21:15:51.3027746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 10%] 2024-08-06T21:15:51.3029019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 10%] 2024-08-06T21:15:51.3030316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 10%] 2024-08-06T21:15:51.3031632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 10%] 2024-08-06T21:15:51.3032929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 10%] 2024-08-06T21:15:51.3034190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 10%] 2024-08-06T21:15:51.3035479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 10%] 2024-08-06T21:15:51.3036736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 10%] 2024-08-06T21:15:51.3038012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 10%] 2024-08-06T21:15:51.3039265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 10%] 2024-08-06T21:15:51.3040543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 10%] 2024-08-06T21:15:51.3041796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 10%] 2024-08-06T21:15:51.3043066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 10%] 2024-08-06T21:15:51.3044382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 10%] 2024-08-06T21:15:51.3045693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 10%] 2024-08-06T21:15:51.3046963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 10%] 2024-08-06T21:15:51.3048225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 10%] 2024-08-06T21:15:51.3049541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 10%] 2024-08-06T21:15:51.3050839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 10%] 2024-08-06T21:15:51.3052101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 10%] 2024-08-06T21:15:51.3053380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 10%] 2024-08-06T21:15:51.3054643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 10%] 2024-08-06T21:15:51.3055925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 10%] 2024-08-06T21:15:51.3057174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 10%] 2024-08-06T21:15:51.3058458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 10%] 2024-08-06T21:15:51.3059714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 10%] 2024-08-06T21:15:51.3060992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 10%] 2024-08-06T21:15:51.3062291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 10%] 2024-08-06T21:15:51.3063620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 10%] 2024-08-06T21:15:51.3064877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 10%] 2024-08-06T21:15:51.3066152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 10%] 2024-08-06T21:15:51.3067446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 10%] 2024-08-06T21:15:51.3068756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 10%] 2024-08-06T21:15:51.3070019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 10%] 2024-08-06T21:15:51.3071266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 10%] 2024-08-06T21:15:51.3072544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 10%] 2024-08-06T21:15:51.3073804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 10%] 2024-08-06T21:15:51.3075068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 10%] 2024-08-06T21:15:51.3076323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 10%] 2024-08-06T21:15:51.3077596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 10%] 2024-08-06T21:15:51.3078856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 10%] 2024-08-06T21:15:51.3080124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0136s] [ 10%] 2024-08-06T21:15:51.3081465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0145s] [ 10%] 2024-08-06T21:15:51.3082808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0083s] [ 10%] 2024-08-06T21:15:51.3084102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0083s] [ 10%] 2024-08-06T21:15:51.3085363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0213s] [ 10%] 2024-08-06T21:15:51.3086755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0217s] [ 11%] 2024-08-06T21:15:51.3088082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0092s] [ 11%] 2024-08-06T21:15:51.3089378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0091s] [ 11%] 2024-08-06T21:15:51.3090651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0140s] [ 11%] 2024-08-06T21:15:51.3091949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0155s] [ 11%] 2024-08-06T21:15:51.3093235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0091s] [ 11%] 2024-08-06T21:15:51.3094503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0085s] [ 11%] 2024-08-06T21:15:51.3095794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0215s] [ 11%] 2024-08-06T21:15:51.3097077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0218s] [ 11%] 2024-08-06T21:15:51.3098369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0092s] [ 11%] 2024-08-06T21:15:51.3099687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0092s] [ 11%] 2024-08-06T21:15:51.3101029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0152s] [ 11%] 2024-08-06T21:15:51.3102363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0169s] [ 11%] 2024-08-06T21:15:51.3103653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0086s] [ 11%] 2024-08-06T21:15:51.3105376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0087s] [ 11%] 2024-08-06T21:15:51.3106776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0225s] [ 11%] 2024-08-06T21:15:51.3108071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0229s] [ 11%] 2024-08-06T21:15:51.3109337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0096s] [ 11%] 2024-08-06T21:15:51.3110629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0096s] [ 11%] 2024-08-06T21:15:51.3111901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0132s] [ 11%] 2024-08-06T21:15:51.3113208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0138s] [ 11%] 2024-08-06T21:15:51.3114479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0080s] [ 11%] 2024-08-06T21:15:51.3115771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 11%] 2024-08-06T21:15:51.3117038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0210s] [ 11%] 2024-08-06T21:15:51.3118323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0214s] [ 11%] 2024-08-06T21:15:51.3119648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0089s] [ 11%] 2024-08-06T21:15:51.3120984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0089s] [ 11%] 2024-08-06T21:15:51.3122266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0238s] [ 11%] 2024-08-06T21:15:51.3123558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0263s] [ 11%] 2024-08-06T21:15:51.3124894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0105s] [ 11%] 2024-08-06T21:15:51.3126216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0106s] [ 11%] 2024-08-06T21:15:51.3127504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0383s] [ 11%] 2024-08-06T21:15:51.3128782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0396s] [ 11%] 2024-08-06T21:15:51.3130080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0119s] [ 11%] 2024-08-06T21:15:51.3131356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0115s] [ 11%] 2024-08-06T21:15:51.3132643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0245s] [ 11%] 2024-08-06T21:15:51.3133940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0277s] [ 11%] 2024-08-06T21:15:51.3135211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0108s] [ 11%] 2024-08-06T21:15:51.3136498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0111s] [ 11%] 2024-08-06T21:15:51.3137812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0387s] [ 11%] 2024-08-06T21:15:51.3139163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0396s] [ 11%] 2024-08-06T21:15:51.3140436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0122s] [ 11%] 2024-08-06T21:15:51.3141726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0124s] [ 11%] 2024-08-06T21:15:51.3143055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0278s] [ 11%] 2024-08-06T21:15:51.3144401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0314s] [ 11%] 2024-08-06T21:15:51.3145666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0120s] [ 11%] 2024-08-06T21:15:51.3146939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0119s] [ 11%] 2024-08-06T21:15:51.3148236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0414s] [ 11%] 2024-08-06T21:15:51.3149516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0417s] [ 11%] 2024-08-06T21:15:51.3150798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0129s] [ 11%] 2024-08-06T21:15:51.3152075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0131s] [ 11%] 2024-08-06T21:15:51.3153382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0229s] [ 11%] 2024-08-06T21:15:51.3154654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0249s] [ 11%] 2024-08-06T21:15:51.3155936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0107s] [ 11%] 2024-08-06T21:15:51.3157242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0102s] [ 11%] 2024-08-06T21:15:51.3158579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0378s] [ 11%] 2024-08-06T21:15:51.3159851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0386s] [ 11%] 2024-08-06T21:15:51.3161158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0114s] [ 11%] 2024-08-06T21:15:51.3162499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0117s] [ 11%] 2024-08-06T21:15:51.3163766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0092s] [ 11%] 2024-08-06T21:15:51.3165050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0096s] [ 11%] 2024-08-06T21:15:51.3166314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0077s] [ 11%] 2024-08-06T21:15:51.3167602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0076s] [ 11%] 2024-08-06T21:15:51.3168863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0137s] [ 11%] 2024-08-06T21:15:51.3170146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0140s] [ 11%] 2024-08-06T21:15:51.3171412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0083s] [ 11%] 2024-08-06T21:15:51.3172702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 11%] 2024-08-06T21:15:51.3173981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0092s] [ 11%] 2024-08-06T21:15:51.3175291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0094s] [ 11%] 2024-08-06T21:15:51.3176639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0073s] [ 11%] 2024-08-06T21:15:51.3177905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 11%] 2024-08-06T21:15:51.3179182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0140s] [ 11%] 2024-08-06T21:15:51.3180493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0139s] [ 11%] 2024-08-06T21:15:51.3181829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0084s] [ 11%] 2024-08-06T21:15:51.3183116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0083s] [ 11%] 2024-08-06T21:15:51.3184397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0105s] [ 11%] 2024-08-06T21:15:51.3185671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 11%] 2024-08-06T21:15:51.3186931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0077s] [ 12%] 2024-08-06T21:15:51.3188204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 12%] 2024-08-06T21:15:51.3189472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0150s] [ 12%] 2024-08-06T21:15:51.3190770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0152s] [ 12%] 2024-08-06T21:15:51.3192038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0083s] [ 12%] 2024-08-06T21:15:51.3193377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0083s] [ 12%] 2024-08-06T21:15:51.3194690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0088s] [ 12%] 2024-08-06T21:15:51.3195977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0090s] [ 12%] 2024-08-06T21:15:51.3197236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0073s] [ 12%] 2024-08-06T21:15:51.3198542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 12%] 2024-08-06T21:15:51.3199868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0131s] [ 12%] 2024-08-06T21:15:51.3201137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0133s] [ 12%] 2024-08-06T21:15:51.3202469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0079s] [ 12%] 2024-08-06T21:15:51.3203745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 12%] 2024-08-06T21:15:51.3205401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0085s] [ 12%] 2024-08-06T21:15:51.3206678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 12%] 2024-08-06T21:15:51.3207954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0072s] [ 12%] 2024-08-06T21:15:51.3209226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 12%] 2024-08-06T21:15:51.3210506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0122s] [ 12%] 2024-08-06T21:15:51.3211771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0122s] [ 12%] 2024-08-06T21:15:51.3213130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0080s] [ 12%] 2024-08-06T21:15:51.3214493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 12%] 2024-08-06T21:15:51.3215753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0087s] [ 12%] 2024-08-06T21:15:51.3217027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0087s] [ 12%] 2024-08-06T21:15:51.3218353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0073s] [ 12%] 2024-08-06T21:15:51.3219701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 12%] 2024-08-06T21:15:51.3220962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0127s] [ 12%] 2024-08-06T21:15:51.3222248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0128s] [ 12%] 2024-08-06T21:15:51.3223537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0082s] [ 12%] 2024-08-06T21:15:51.3224806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0081s] [ 12%] 2024-08-06T21:15:51.3226075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0098s] [ 12%] 2024-08-06T21:15:51.3227340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0102s] [ 12%] 2024-08-06T21:15:51.3228623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0078s] [ 12%] 2024-08-06T21:15:51.3229883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 12%] 2024-08-06T21:15:51.3231206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0138s] [ 12%] 2024-08-06T21:15:51.3232520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0139s] [ 12%] 2024-08-06T21:15:51.3233802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0080s] [ 12%] 2024-08-06T21:15:51.3235070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 12%] 2024-08-06T21:15:51.3236389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0082s] [ 12%] 2024-08-06T21:15:51.3237701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 12%] 2024-08-06T21:15:51.3238961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0074s] [ 12%] 2024-08-06T21:15:51.3240238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 12%] 2024-08-06T21:15:51.3241506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0122s] [ 12%] 2024-08-06T21:15:51.3242817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0120s] [ 12%] 2024-08-06T21:15:51.3244078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0077s] [ 12%] 2024-08-06T21:15:51.3245362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0076s] [ 12%] 2024-08-06T21:15:51.3246638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0457s] [ 12%] 2024-08-06T21:15:51.3247930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0516s] [ 12%] 2024-08-06T21:15:51.3249195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0151s] [ 12%] 2024-08-06T21:15:51.3250525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0147s] [ 12%] 2024-08-06T21:15:51.3251869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0730s] [ 12%] 2024-08-06T21:15:51.3253174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0768s] [ 12%] 2024-08-06T21:15:51.3254462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0172s] [ 12%] 2024-08-06T21:15:51.3255784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0172s] [ 12%] 2024-08-06T21:15:51.3257130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0473s] [ 12%] 2024-08-06T21:15:51.3258403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0541s] [ 12%] 2024-08-06T21:15:51.3259685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0155s] [ 12%] 2024-08-06T21:15:51.3260967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0160s] [ 12%] 2024-08-06T21:15:51.3262250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0733s] [ 12%] 2024-08-06T21:15:51.3263537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0769s] [ 12%] 2024-08-06T21:15:51.3264808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0179s] [ 12%] 2024-08-06T21:15:51.3266108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0181s] [ 12%] 2024-08-06T21:15:51.3267375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0529s] [ 12%] 2024-08-06T21:15:51.3268703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0599s] [ 12%] 2024-08-06T21:15:51.3270020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0174s] [ 12%] 2024-08-06T21:15:51.3271310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0173s] [ 12%] 2024-08-06T21:15:51.3272590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0774s] [ 12%] 2024-08-06T21:15:51.3273923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0802s] [ 12%] 2024-08-06T21:15:51.3275244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0193s] [ 12%] 2024-08-06T21:15:51.3276521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0200s] [ 12%] 2024-08-06T21:15:51.3277803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0441s] [ 12%] 2024-08-06T21:15:51.3279077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0485s] [ 12%] 2024-08-06T21:15:51.3280359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0146s] [ 12%] 2024-08-06T21:15:51.3281643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0144s] [ 12%] 2024-08-06T21:15:51.3282959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0721s] [ 12%] 2024-08-06T21:15:51.3284241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0752s] [ 12%] 2024-08-06T21:15:51.3285521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0167s] [ 12%] 2024-08-06T21:15:51.3286789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0169s] [ 12%] 2024-08-06T21:15:51.3288110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0092s] [ 13%] 2024-08-06T21:15:51.3289428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0096s] [ 13%] 2024-08-06T21:15:51.3290684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0074s] [ 13%] 2024-08-06T21:15:51.3291965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 13%] 2024-08-06T21:15:51.3293284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0133s] [ 13%] 2024-08-06T21:15:51.3294612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0137s] [ 13%] 2024-08-06T21:15:51.3295878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0083s] [ 13%] 2024-08-06T21:15:51.3297172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 13%] 2024-08-06T21:15:51.3298440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0095s] [ 13%] 2024-08-06T21:15:51.3299708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0098s] [ 13%] 2024-08-06T21:15:51.3300965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0078s] [ 13%] 2024-08-06T21:15:51.3302286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0077s] [ 13%] 2024-08-06T21:15:51.3303576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0137s] [ 13%] 2024-08-06T21:15:51.3305250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0137s] [ 13%] 2024-08-06T21:15:51.3306629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0083s] [ 13%] 2024-08-06T21:15:51.3307974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 13%] 2024-08-06T21:15:51.3309248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0107s] [ 13%] 2024-08-06T21:15:51.3310511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0111s] [ 13%] 2024-08-06T21:15:51.3311841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0065s] [ 13%] 2024-08-06T21:15:51.3313208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 13%] 2024-08-06T21:15:51.3314495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0147s] [ 13%] 2024-08-06T21:15:51.3315772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0146s] [ 13%] 2024-08-06T21:15:51.3317041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0088s] [ 13%] 2024-08-06T21:15:51.3318330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0088s] [ 13%] 2024-08-06T21:15:51.3319592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0092s] [ 13%] 2024-08-06T21:15:51.3320875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0093s] [ 13%] 2024-08-06T21:15:51.3322140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0076s] [ 13%] 2024-08-06T21:15:51.3323430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 13%] 2024-08-06T21:15:51.3324691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0134s] [ 13%] 2024-08-06T21:15:51.3326013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0134s] [ 13%] 2024-08-06T21:15:51.3327322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0081s] [ 13%] 2024-08-06T21:15:51.3328587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0083s] [ 13%] 2024-08-06T21:15:51.3329858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0087s] [ 13%] 2024-08-06T21:15:51.3331161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0086s] [ 13%] 2024-08-06T21:15:51.3332479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0073s] [ 13%] 2024-08-06T21:15:51.3333744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 13%] 2024-08-06T21:15:51.3335018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0126s] [ 13%] 2024-08-06T21:15:51.3336288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0127s] [ 13%] 2024-08-06T21:15:51.3337551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0079s] [ 13%] 2024-08-06T21:15:51.3338810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 13%] 2024-08-06T21:15:51.3340088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0090s] [ 13%] 2024-08-06T21:15:51.3341363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0092s] [ 13%] 2024-08-06T21:15:51.3342643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0069s] [ 13%] 2024-08-06T21:15:51.3343961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 13%] 2024-08-06T21:15:51.3345273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0131s] [ 13%] 2024-08-06T21:15:51.3346557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0132s] [ 13%] 2024-08-06T21:15:51.3347812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0084s] [ 13%] 2024-08-06T21:15:51.3349125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 13%] 2024-08-06T21:15:51.3350436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0103s] [ 13%] 2024-08-06T21:15:51.3351716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0104s] [ 13%] 2024-08-06T21:15:51.3353001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0079s] [ 13%] 2024-08-06T21:15:51.3354271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 13%] 2024-08-06T21:15:51.3355539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0144s] [ 13%] 2024-08-06T21:15:51.3356804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0146s] [ 13%] 2024-08-06T21:15:51.3358080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0084s] [ 13%] 2024-08-06T21:15:51.3359344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 13%] 2024-08-06T21:15:51.3360611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0084s] [ 13%] 2024-08-06T21:15:51.3361868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 13%] 2024-08-06T21:15:51.3363198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0071s] [ 13%] 2024-08-06T21:15:51.3364513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 13%] 2024-08-06T21:15:51.3365786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0125s] [ 13%] 2024-08-06T21:15:51.3367059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0125s] [ 13%] 2024-08-06T21:15:51.3368363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0080s] [ 13%] 2024-08-06T21:15:51.3369691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 13%] 2024-08-06T21:15:51.3370952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 13%] 2024-08-06T21:15:51.3372258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 13%] 2024-08-06T21:15:51.3373530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 13%] 2024-08-06T21:15:51.3374813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 13%] 2024-08-06T21:15:51.3376079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0072s] [ 13%] 2024-08-06T21:15:51.3377372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0070s] [ 13%] 2024-08-06T21:15:51.3378643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 13%] 2024-08-06T21:15:51.3379909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 13%] 2024-08-06T21:15:51.3381231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 13%] 2024-08-06T21:15:51.3382642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 13%] 2024-08-06T21:15:51.3383917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 13%] 2024-08-06T21:15:51.3385180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 13%] 2024-08-06T21:15:51.3386517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0073s] [ 13%] 2024-08-06T21:15:51.3387838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 14%] 2024-08-06T21:15:51.3389113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 14%] 2024-08-06T21:15:51.3390380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 14%] 2024-08-06T21:15:51.3391676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0062s] [ 14%] 2024-08-06T21:15:51.3392963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 14%] 2024-08-06T21:15:51.3394220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 14%] 2024-08-06T21:15:51.3395495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 14%] 2024-08-06T21:15:51.3396765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0075s] [ 14%] 2024-08-06T21:15:51.3398038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 14%] 2024-08-06T21:15:51.3399296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 14%] 2024-08-06T21:15:51.3400623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 14%] 2024-08-06T21:15:51.3401936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 14%] 2024-08-06T21:15:51.3403260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 14%] 2024-08-06T21:15:51.3404909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 14%] 2024-08-06T21:15:51.3406278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 14%] 2024-08-06T21:15:51.3407652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0072s] [ 14%] 2024-08-06T21:15:51.3408922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 14%] 2024-08-06T21:15:51.3410191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 14%] 2024-08-06T21:15:51.3411464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 14%] 2024-08-06T21:15:51.3412749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0074s] [ 14%] 2024-08-06T21:15:51.3414018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 14%] 2024-08-06T21:15:51.3415304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 14%] 2024-08-06T21:15:51.3416571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 14%] 2024-08-06T21:15:51.3417846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0095s] [ 14%] 2024-08-06T21:15:51.3419175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0096s] [ 14%] 2024-08-06T21:15:51.3420505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 14%] 2024-08-06T21:15:51.3421784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 14%] 2024-08-06T21:15:51.3423083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0075s] [ 14%] 2024-08-06T21:15:51.3424414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 14%] 2024-08-06T21:15:51.3425731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 14%] 2024-08-06T21:15:51.3427008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 14%] 2024-08-06T21:15:51.3428272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0099s] [ 14%] 2024-08-06T21:15:51.3429565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0098s] [ 14%] 2024-08-06T21:15:51.3430832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 14%] 2024-08-06T21:15:51.3432098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 14%] 2024-08-06T21:15:51.3433380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0082s] [ 14%] 2024-08-06T21:15:51.3434647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0088s] [ 14%] 2024-08-06T21:15:51.3435911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0064s] [ 14%] 2024-08-06T21:15:51.3437171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 14%] 2024-08-06T21:15:51.3438486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0103s] [ 14%] 2024-08-06T21:15:51.3439797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 14%] 2024-08-06T21:15:51.3441068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 14%] 2024-08-06T21:15:51.3442389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 14%] 2024-08-06T21:15:51.3443708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0074s] [ 14%] 2024-08-06T21:15:51.3444971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 14%] 2024-08-06T21:15:51.3446224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 14%] 2024-08-06T21:15:51.3447506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 14%] 2024-08-06T21:15:51.3448774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0095s] [ 14%] 2024-08-06T21:15:51.3450056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0096s] [ 14%] 2024-08-06T21:15:51.3451317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 14%] 2024-08-06T21:15:51.3452622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 14%] 2024-08-06T21:15:51.3453878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 14%] 2024-08-06T21:15:51.3455145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 14%] 2024-08-06T21:15:51.3456441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 14%] 2024-08-06T21:15:51.3457771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 14%] 2024-08-06T21:15:51.3459043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 14%] 2024-08-06T21:15:51.3460307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 14%] 2024-08-06T21:15:51.3461652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 14%] 2024-08-06T21:15:51.3463002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 14%] 2024-08-06T21:15:51.3464270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 14%] 2024-08-06T21:15:51.3465533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 14%] 2024-08-06T21:15:51.3466823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 14%] 2024-08-06T21:15:51.3468090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 14%] 2024-08-06T21:15:51.3469369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 14%] 2024-08-06T21:15:51.3470641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 14%] 2024-08-06T21:15:51.3471910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 14%] 2024-08-06T21:15:51.3473204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 14%] 2024-08-06T21:15:51.3474462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 14%] 2024-08-06T21:15:51.3475793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 14%] 2024-08-06T21:15:51.3477106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 14%] 2024-08-06T21:15:51.3478391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 14%] 2024-08-06T21:15:51.3479692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 14%] 2024-08-06T21:15:51.3481031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 14%] 2024-08-06T21:15:51.3482308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 14%] 2024-08-06T21:15:51.3483576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 14%] 2024-08-06T21:15:51.3484857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 14%] 2024-08-06T21:15:51.3486126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 14%] 2024-08-06T21:15:51.3487400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 15%] 2024-08-06T21:15:51.3488664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 15%] 2024-08-06T21:15:51.3489946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 15%] 2024-08-06T21:15:51.3491208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 15%] 2024-08-06T21:15:51.3492501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 15%] 2024-08-06T21:15:51.3493824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 15%] 2024-08-06T21:15:51.3495164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 15%] 2024-08-06T21:15:51.3496425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 15%] 2024-08-06T21:15:51.3497676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 15%] 2024-08-06T21:15:51.3498997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 15%] 2024-08-06T21:15:51.3500307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 15%] 2024-08-06T21:15:51.3501588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 15%] 2024-08-06T21:15:51.3502906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 15%] 2024-08-06T21:15:51.3504580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 15%] 2024-08-06T21:15:51.3505873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 15%] 2024-08-06T21:15:51.3507150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 15%] 2024-08-06T21:15:51.3508408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 15%] 2024-08-06T21:15:51.3509677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 15%] 2024-08-06T21:15:51.3510953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 15%] 2024-08-06T21:15:51.3512216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 15%] 2024-08-06T21:15:51.3513586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 15%] 2024-08-06T21:15:51.3514936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 15%] 2024-08-06T21:15:51.3516212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 15%] 2024-08-06T21:15:51.3517477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 15%] 2024-08-06T21:15:51.3518803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 15%] 2024-08-06T21:15:51.3520121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 15%] 2024-08-06T21:15:51.3521380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 15%] 2024-08-06T21:15:51.3522677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 15%] 2024-08-06T21:15:51.3523942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 15%] 2024-08-06T21:15:51.3525225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 15%] 2024-08-06T21:15:51.3526473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 15%] 2024-08-06T21:15:51.3527750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 15%] 2024-08-06T21:15:51.3529003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 15%] 2024-08-06T21:15:51.3530276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 15%] 2024-08-06T21:15:51.3531589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0068s] [ 15%] 2024-08-06T21:15:51.3532946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 15%] 2024-08-06T21:15:51.3534199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 15%] 2024-08-06T21:15:51.3535461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 15%] 2024-08-06T21:15:51.3536787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0110s] [ 15%] 2024-08-06T21:15:51.3538113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0117s] [ 15%] 2024-08-06T21:15:51.3539387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0072s] [ 15%] 2024-08-06T21:15:51.3540656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 15%] 2024-08-06T21:15:51.3541946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0149s] [ 15%] 2024-08-06T21:15:51.3543236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0154s] [ 15%] 2024-08-06T21:15:51.3544523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0078s] [ 15%] 2024-08-06T21:15:51.3545801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 15%] 2024-08-06T21:15:51.3547068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0111s] [ 15%] 2024-08-06T21:15:51.3548350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0121s] [ 15%] 2024-08-06T21:15:51.3549609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0075s] [ 15%] 2024-08-06T21:15:51.3550937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0076s] [ 15%] 2024-08-06T21:15:51.3552255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0153s] [ 15%] 2024-08-06T21:15:51.3553543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0156s] [ 15%] 2024-08-06T21:15:51.3554811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0081s] [ 15%] 2024-08-06T21:15:51.3556144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0081s] [ 15%] 2024-08-06T21:15:51.3557456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0125s] [ 15%] 2024-08-06T21:15:51.3558742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0129s] [ 15%] 2024-08-06T21:15:51.3560007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0074s] [ 15%] 2024-08-06T21:15:51.3561279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 15%] 2024-08-06T21:15:51.3562584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0157s] [ 15%] 2024-08-06T21:15:51.3563860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0159s] [ 15%] 2024-08-06T21:15:51.3565150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0084s] [ 15%] 2024-08-06T21:15:51.3566423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0083s] [ 15%] 2024-08-06T21:15:51.3567700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0105s] [ 15%] 2024-08-06T21:15:51.3569009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0111s] [ 15%] 2024-08-06T21:15:51.3570344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0071s] [ 15%] 2024-08-06T21:15:51.3571608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 15%] 2024-08-06T21:15:51.3572893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0144s] [ 15%] 2024-08-06T21:15:51.3574226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0152s] [ 15%] 2024-08-06T21:15:51.3575541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0077s] [ 15%] 2024-08-06T21:15:51.3576829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 15%] 2024-08-06T21:15:51.3578092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 15%] 2024-08-06T21:15:51.3579382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 15%] 2024-08-06T21:15:51.3580645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 15%] 2024-08-06T21:15:51.3581930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 15%] 2024-08-06T21:15:51.3583209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 15%] 2024-08-06T21:15:51.3584505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 15%] 2024-08-06T21:15:51.3585768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 15%] 2024-08-06T21:15:51.3587030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 15%] 2024-08-06T21:15:51.3588356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 16%] 2024-08-06T21:15:51.3589675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 16%] 2024-08-06T21:15:51.3590956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 16%] 2024-08-06T21:15:51.3592269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 16%] 2024-08-06T21:15:51.3593619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 16%] 2024-08-06T21:15:51.3594884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 16%] 2024-08-06T21:15:51.3596160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 16%] 2024-08-06T21:15:51.3597433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 16%] 2024-08-06T21:15:51.3598703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 16%] 2024-08-06T21:15:51.3599984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 16%] 2024-08-06T21:15:51.3601242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 16%] 2024-08-06T21:15:51.3602593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 16%] 2024-08-06T21:15:51.3603862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 16%] 2024-08-06T21:15:51.3605520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 16%] 2024-08-06T21:15:51.3606864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 16%] 2024-08-06T21:15:51.3608220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 16%] 2024-08-06T21:15:51.3609473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 16%] 2024-08-06T21:15:51.3610746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 16%] 2024-08-06T21:15:51.3612060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 16%] 2024-08-06T21:15:51.3613405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 16%] 2024-08-06T21:15:51.3614685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 16%] 2024-08-06T21:15:51.3615956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 16%] 2024-08-06T21:15:51.3617242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 16%] 2024-08-06T21:15:51.3618509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 16%] 2024-08-06T21:15:51.3619785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 16%] 2024-08-06T21:15:51.3621055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 16%] 2024-08-06T21:15:51.3622333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 16%] 2024-08-06T21:15:51.3623610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 16%] 2024-08-06T21:15:51.3624869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 16%] 2024-08-06T21:15:51.3626197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 16%] 2024-08-06T21:15:51.3627507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 16%] 2024-08-06T21:15:51.3628792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 16%] 2024-08-06T21:15:51.3630051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 16%] 2024-08-06T21:15:51.3631375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 16%] 2024-08-06T21:15:51.3632689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 16%] 2024-08-06T21:15:51.3633967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 16%] 2024-08-06T21:15:51.3635229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 16%] 2024-08-06T21:15:51.3636520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 16%] 2024-08-06T21:15:51.3637777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 16%] 2024-08-06T21:15:51.3639039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 16%] 2024-08-06T21:15:51.3640318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 16%] 2024-08-06T21:15:51.3641584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 16%] 2024-08-06T21:15:51.3642883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 16%] 2024-08-06T21:15:51.3644190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 16%] 2024-08-06T21:15:51.3645519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 16%] 2024-08-06T21:15:51.3646781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 16%] 2024-08-06T21:15:51.3648054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 16%] 2024-08-06T21:15:51.3649362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 16%] 2024-08-06T21:15:51.3650667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 16%] 2024-08-06T21:15:51.3651944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 16%] 2024-08-06T21:15:51.3653218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 16%] 2024-08-06T21:15:51.3654500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 16%] 2024-08-06T21:15:51.3655762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 16%] 2024-08-06T21:15:51.3657042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 16%] 2024-08-06T21:15:51.3658303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 16%] 2024-08-06T21:15:51.3659589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 16%] 2024-08-06T21:15:51.3660850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 16%] 2024-08-06T21:15:51.3662117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 16%] 2024-08-06T21:15:51.3663457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 16%] 2024-08-06T21:15:51.3664781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 16%] 2024-08-06T21:15:51.3666063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 16%] 2024-08-06T21:15:51.3667336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 16%] 2024-08-06T21:15:51.3668681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 16%] 2024-08-06T21:15:51.3670000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 16%] 2024-08-06T21:15:51.3671284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 16%] 2024-08-06T21:15:51.3672563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 16%] 2024-08-06T21:15:51.3673852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 16%] 2024-08-06T21:15:51.3675119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 16%] 2024-08-06T21:15:51.3676383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 16%] 2024-08-06T21:15:51.3677677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 16%] 2024-08-06T21:15:51.3678947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 16%] 2024-08-06T21:15:51.3680229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 16%] 2024-08-06T21:15:51.3681598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 16%] 2024-08-06T21:15:51.3682959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 16%] 2024-08-06T21:15:51.3684226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 16%] 2024-08-06T21:15:51.3685501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 16%] 2024-08-06T21:15:51.3686812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 16%] 2024-08-06T21:15:51.3688135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 17%] 2024-08-06T21:15:51.3689415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 17%] 2024-08-06T21:15:51.3690684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 17%] 2024-08-06T21:15:51.3691965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 17%] 2024-08-06T21:15:51.3693256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 17%] 2024-08-06T21:15:51.3694534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 17%] 2024-08-06T21:15:51.3695802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 17%] 2024-08-06T21:15:51.3697088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 17%] 2024-08-06T21:15:51.3698356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 17%] 2024-08-06T21:15:51.3699632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 17%] 2024-08-06T21:15:51.3700946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 17%] 2024-08-06T21:15:51.3702303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 17%] 2024-08-06T21:15:51.3703615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 17%] 2024-08-06T21:15:51.3705243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 17%] 2024-08-06T21:15:51.3706637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 17%] 2024-08-06T21:15:51.3708017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0072s] [ 17%] 2024-08-06T21:15:51.3709304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 17%] 2024-08-06T21:15:51.3710572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 17%] 2024-08-06T21:15:51.3711871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 17%] 2024-08-06T21:15:51.3713144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 17%] 2024-08-06T21:15:51.3714411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 17%] 2024-08-06T21:15:51.3715698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 17%] 2024-08-06T21:15:51.3716967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 17%] 2024-08-06T21:15:51.3718246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0073s] [ 17%] 2024-08-06T21:15:51.3719568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 17%] 2024-08-06T21:15:51.3720925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 17%] 2024-08-06T21:15:51.3722196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 17%] 2024-08-06T21:15:51.3723492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0075s] [ 17%] 2024-08-06T21:15:51.3724796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 17%] 2024-08-06T21:15:51.3726129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 17%] 2024-08-06T21:15:51.3727395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 17%] 2024-08-06T21:15:51.3728654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0080s] [ 17%] 2024-08-06T21:15:51.3729947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 17%] 2024-08-06T21:15:51.3731214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 17%] 2024-08-06T21:15:51.3732497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 17%] 2024-08-06T21:15:51.3733775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0062s] [ 17%] 2024-08-06T21:15:51.3735061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 17%] 2024-08-06T21:15:51.3736314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 17%] 2024-08-06T21:15:51.3737589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 17%] 2024-08-06T21:15:51.3738894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0068s] [ 17%] 2024-08-06T21:15:51.3740232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 17%] 2024-08-06T21:15:51.3741509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 17%] 2024-08-06T21:15:51.3742775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 17%] 2024-08-06T21:15:51.3744109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 17%] 2024-08-06T21:15:51.3745416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 17%] 2024-08-06T21:15:51.3746688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 17%] 2024-08-06T21:15:51.3747953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 17%] 2024-08-06T21:15:51.3749237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 17%] 2024-08-06T21:15:51.3750501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 17%] 2024-08-06T21:15:51.3751774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 17%] 2024-08-06T21:15:51.3753056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 17%] 2024-08-06T21:15:51.3754320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 17%] 2024-08-06T21:15:51.3755603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 17%] 2024-08-06T21:15:51.3756900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 17%] 2024-08-06T21:15:51.3758227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 17%] 2024-08-06T21:15:51.3759487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 17%] 2024-08-06T21:15:51.3760766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 17%] 2024-08-06T21:15:51.3762070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 17%] 2024-08-06T21:15:51.3763426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 17%] 2024-08-06T21:15:51.3764682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 17%] 2024-08-06T21:15:51.3765938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 17%] 2024-08-06T21:15:51.3767216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 17%] 2024-08-06T21:15:51.3768482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 17%] 2024-08-06T21:15:51.3769753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 17%] 2024-08-06T21:15:51.3771016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 17%] 2024-08-06T21:15:51.3772303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 17%] 2024-08-06T21:15:51.3773585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 17%] 2024-08-06T21:15:51.3774851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 17%] 2024-08-06T21:15:51.3776147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 17%] 2024-08-06T21:15:51.3777472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 17%] 2024-08-06T21:15:51.3778725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 17%] 2024-08-06T21:15:51.3779977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0068s] [ 17%] 2024-08-06T21:15:51.3781296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 17%] 2024-08-06T21:15:51.3782602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 17%] 2024-08-06T21:15:51.3783892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 17%] 2024-08-06T21:15:51.3785151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 17%] 2024-08-06T21:15:51.3786430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 17%] 2024-08-06T21:15:51.3787679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 17%] 2024-08-06T21:15:51.3788951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 18%] 2024-08-06T21:15:51.3790212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 18%] 2024-08-06T21:15:51.3791478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 18%] 2024-08-06T21:15:51.3792758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 18%] 2024-08-06T21:15:51.3794019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 18%] 2024-08-06T21:15:51.3795331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 18%] 2024-08-06T21:15:51.3796643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 18%] 2024-08-06T21:15:51.3797910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 18%] 2024-08-06T21:15:51.3799204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 18%] 2024-08-06T21:15:51.3800534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 18%] 2024-08-06T21:15:51.3801794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 18%] 2024-08-06T21:15:51.3803103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 18%] 2024-08-06T21:15:51.3804755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 18%] 2024-08-06T21:15:51.3806044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 18%] 2024-08-06T21:15:51.3807325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 18%] 2024-08-06T21:15:51.3808576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 18%] 2024-08-06T21:15:51.3809856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 18%] 2024-08-06T21:15:51.3811116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 18%] 2024-08-06T21:15:51.3812393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 18%] 2024-08-06T21:15:51.3813765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 18%] 2024-08-06T21:15:51.3815122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 18%] 2024-08-06T21:15:51.3816372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 18%] 2024-08-06T21:15:51.3817627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 18%] 2024-08-06T21:15:51.3818951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 18%] 2024-08-06T21:15:51.3820276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 18%] 2024-08-06T21:15:51.3821549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 18%] 2024-08-06T21:15:51.3822814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 18%] 2024-08-06T21:15:51.3824110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 18%] 2024-08-06T21:15:51.3825376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 18%] 2024-08-06T21:15:51.3826654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0088s] [ 18%] 2024-08-06T21:15:51.3827927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0091s] [ 18%] 2024-08-06T21:15:51.3829201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 18%] 2024-08-06T21:15:51.3830484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 18%] 2024-08-06T21:15:51.3831747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0098s] [ 18%] 2024-08-06T21:15:51.3833079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0101s] [ 18%] 2024-08-06T21:15:51.3834390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 18%] 2024-08-06T21:15:51.3835677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 18%] 2024-08-06T21:15:51.3836944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0096s] [ 18%] 2024-08-06T21:15:51.3838278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0097s] [ 18%] 2024-08-06T21:15:51.3839585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 18%] 2024-08-06T21:15:51.3840868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 18%] 2024-08-06T21:15:51.3842136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0107s] [ 18%] 2024-08-06T21:15:51.3843429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 18%] 2024-08-06T21:15:51.3844714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 18%] 2024-08-06T21:15:51.3845984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 18%] 2024-08-06T21:15:51.3847272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0113s] [ 18%] 2024-08-06T21:15:51.3848546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0115s] [ 18%] 2024-08-06T21:15:51.3849819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 18%] 2024-08-06T21:15:51.3851129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 18%] 2024-08-06T21:15:51.3852464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0122s] [ 18%] 2024-08-06T21:15:51.3853756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0124s] [ 18%] 2024-08-06T21:15:51.3855024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 18%] 2024-08-06T21:15:51.3856356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 18%] 2024-08-06T21:15:51.3857666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0081s] [ 18%] 2024-08-06T21:15:51.3858948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0083s] [ 18%] 2024-08-06T21:15:51.3860204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 18%] 2024-08-06T21:15:51.3861496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 18%] 2024-08-06T21:15:51.3862760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0092s] [ 18%] 2024-08-06T21:15:51.3864057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0094s] [ 18%] 2024-08-06T21:15:51.3865323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 18%] 2024-08-06T21:15:51.3866595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 18%] 2024-08-06T21:15:51.3867867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 18%] 2024-08-06T21:15:51.3869125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 18%] 2024-08-06T21:15:51.3870442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 18%] 2024-08-06T21:15:51.3871756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 18%] 2024-08-06T21:15:51.3873038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 18%] 2024-08-06T21:15:51.3874306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 18%] 2024-08-06T21:15:51.3875634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 18%] 2024-08-06T21:15:51.3876962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 18%] 2024-08-06T21:15:51.3878236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 18%] 2024-08-06T21:15:51.3879501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 18%] 2024-08-06T21:15:51.3880763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 18%] 2024-08-06T21:15:51.3882038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 18%] 2024-08-06T21:15:51.3883314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 18%] 2024-08-06T21:15:51.3884603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 18%] 2024-08-06T21:15:51.3885866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 18%] 2024-08-06T21:15:51.3887151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 18%] 2024-08-06T21:15:51.3888448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 19%] 2024-08-06T21:15:51.3889776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 19%] 2024-08-06T21:15:51.3891035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 19%] 2024-08-06T21:15:51.3892295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 19%] 2024-08-06T21:15:51.3893642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 19%] 2024-08-06T21:15:51.3894965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 19%] 2024-08-06T21:15:51.3896244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 19%] 2024-08-06T21:15:51.3897506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 19%] 2024-08-06T21:15:51.3898777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 19%] 2024-08-06T21:15:51.3900044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 19%] 2024-08-06T21:15:51.3901313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 19%] 2024-08-06T21:15:51.3902613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 19%] 2024-08-06T21:15:51.3904195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 19%] 2024-08-06T21:15:51.3905547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 19%] 2024-08-06T21:15:51.3906806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 19%] 2024-08-06T21:15:51.3908168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 19%] 2024-08-06T21:15:51.3909489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 19%] 2024-08-06T21:15:51.3910762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 19%] 2024-08-06T21:15:51.3912011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 19%] 2024-08-06T21:15:51.3913340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 19%] 2024-08-06T21:15:51.3914676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 19%] 2024-08-06T21:15:51.3915955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 19%] 2024-08-06T21:15:51.3917216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 19%] 2024-08-06T21:15:51.3918487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 19%] 2024-08-06T21:15:51.3919752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 19%] 2024-08-06T21:15:51.3921013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 19%] 2024-08-06T21:15:51.3922293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 19%] 2024-08-06T21:15:51.3923566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 19%] 2024-08-06T21:15:51.3924835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 19%] 2024-08-06T21:15:51.3926140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 19%] 2024-08-06T21:15:51.3927466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 19%] 2024-08-06T21:15:51.3928728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 19%] 2024-08-06T21:15:51.3929979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 19%] 2024-08-06T21:15:51.3931300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 19%] 2024-08-06T21:15:51.3932605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 19%] 2024-08-06T21:15:51.3933901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 19%] 2024-08-06T21:15:51.3935157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 19%] 2024-08-06T21:15:51.3936440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 19%] 2024-08-06T21:15:51.3937703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 19%] 2024-08-06T21:15:51.3938975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 19%] 2024-08-06T21:15:51.3940224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 19%] 2024-08-06T21:15:51.3941498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 19%] 2024-08-06T21:15:51.3942748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 19%] 2024-08-06T21:15:51.3944022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 19%] 2024-08-06T21:15:51.3945334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 19%] 2024-08-06T21:15:51.3946647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 19%] 2024-08-06T21:15:51.3947915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 19%] 2024-08-06T21:15:51.3949173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 19%] 2024-08-06T21:15:51.3950522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0082s] [ 19%] 2024-08-06T21:15:51.3951853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 19%] 2024-08-06T21:15:51.3953140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0079s] [ 19%] 2024-08-06T21:15:51.3954435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 19%] 2024-08-06T21:15:51.3955714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0105s] [ 19%] 2024-08-06T21:15:51.3957014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 19%] 2024-08-06T21:15:51.3958282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0091s] [ 19%] 2024-08-06T21:15:51.3959583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0091s] [ 19%] 2024-08-06T21:15:51.3960860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0087s] [ 19%] 2024-08-06T21:15:51.3962146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0089s] [ 19%] 2024-08-06T21:15:51.3963422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0087s] [ 19%] 2024-08-06T21:15:51.3964762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0088s] [ 19%] 2024-08-06T21:15:51.3966086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0111s] [ 19%] 2024-08-06T21:15:51.3967378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0112s] [ 19%] 2024-08-06T21:15:51.3968693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0101s] [ 19%] 2024-08-06T21:15:51.3970101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0100s] [ 19%] 2024-08-06T21:15:51.3971384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0093s] [ 19%] 2024-08-06T21:15:51.3972658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0099s] [ 19%] 2024-08-06T21:15:51.3973977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0100s] [ 19%] 2024-08-06T21:15:51.3975262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0100s] [ 19%] 2024-08-06T21:15:51.3976549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0119s] [ 19%] 2024-08-06T21:15:51.3977825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0124s] [ 19%] 2024-08-06T21:15:51.3979116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0111s] [ 19%] 2024-08-06T21:15:51.3980394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0112s] [ 19%] 2024-08-06T21:15:51.3981653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0081s] [ 19%] 2024-08-06T21:15:51.3982984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 19%] 2024-08-06T21:15:51.3984320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0079s] [ 19%] 2024-08-06T21:15:51.3985612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 19%] 2024-08-06T21:15:51.3986883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0106s] [ 19%] 2024-08-06T21:15:51.3988222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0107s] [ 20%] 2024-08-06T21:15:51.3989544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0087s] [ 20%] 2024-08-06T21:15:51.3990836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0088s] [ 20%] 2024-08-06T21:15:51.3992113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0123s] [ 20%] 2024-08-06T21:15:51.3993420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0132s] [ 20%] 2024-08-06T21:15:51.3994685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0118s] [ 20%] 2024-08-06T21:15:51.3995958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0121s] [ 20%] 2024-08-06T21:15:51.3997254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0175s] [ 20%] 2024-08-06T21:15:51.3998533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0181s] [ 20%] 2024-08-06T21:15:51.3999823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0137s] [ 20%] 2024-08-06T21:15:51.4001149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0138s] [ 20%] 2024-08-06T21:15:51.4002533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0131s] [ 20%] 2024-08-06T21:15:51.4003840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0140s] [ 20%] 2024-08-06T21:15:51.4005534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0131s] [ 20%] 2024-08-06T21:15:51.4006928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0134s] [ 20%] 2024-08-06T21:15:51.4008299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0181s] [ 20%] 2024-08-06T21:15:51.4009581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0186s] [ 20%] 2024-08-06T21:15:51.4010856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0151s] [ 20%] 2024-08-06T21:15:51.4012154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0151s] [ 20%] 2024-08-06T21:15:51.4013433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0149s] [ 20%] 2024-08-06T21:15:51.4014745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0158s] [ 20%] 2024-08-06T21:15:51.4016017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0158s] [ 20%] 2024-08-06T21:15:51.4017312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0159s] [ 20%] 2024-08-06T21:15:51.4018585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0200s] [ 20%] 2024-08-06T21:15:51.4019878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0204s] [ 20%] 2024-08-06T21:15:51.4021214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0173s] [ 20%] 2024-08-06T21:15:51.4022585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0175s] [ 20%] 2024-08-06T21:15:51.4023880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0119s] [ 20%] 2024-08-06T21:15:51.4025161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0123s] [ 20%] 2024-08-06T21:15:51.4026492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0115s] [ 20%] 2024-08-06T21:15:51.4027818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0115s] [ 20%] 2024-08-06T21:15:51.4029108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0172s] [ 20%] 2024-08-06T21:15:51.4030391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0174s] [ 20%] 2024-08-06T21:15:51.4031680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0131s] [ 20%] 2024-08-06T21:15:51.4032956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0130s] [ 20%] 2024-08-06T21:15:51.4034263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 20%] 2024-08-06T21:15:51.4035545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 20%] 2024-08-06T21:15:51.4036819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 20%] 2024-08-06T21:15:51.4038110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 20%] 2024-08-06T21:15:51.4039427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 20%] 2024-08-06T21:15:51.4040772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 20%] 2024-08-06T21:15:51.4042043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 20%] 2024-08-06T21:15:51.4043336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 20%] 2024-08-06T21:15:51.4044662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 20%] 2024-08-06T21:15:51.4046010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 20%] 2024-08-06T21:15:51.4047279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0068s] [ 20%] 2024-08-06T21:15:51.4048558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 20%] 2024-08-06T21:15:51.4049857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0082s] [ 20%] 2024-08-06T21:15:51.4051140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0081s] [ 20%] 2024-08-06T21:15:51.4052425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 20%] 2024-08-06T21:15:51.4053703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 20%] 2024-08-06T21:15:51.4054998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0069s] [ 20%] 2024-08-06T21:15:51.4056270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 20%] 2024-08-06T21:15:51.4057546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0068s] [ 20%] 2024-08-06T21:15:51.4058859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 20%] 2024-08-06T21:15:51.4060202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0080s] [ 20%] 2024-08-06T21:15:51.4061480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0081s] [ 20%] 2024-08-06T21:15:51.4062754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0077s] [ 20%] 2024-08-06T21:15:51.4064109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 20%] 2024-08-06T21:15:51.4065427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 20%] 2024-08-06T21:15:51.4066717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 20%] 2024-08-06T21:15:51.4067987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 20%] 2024-08-06T21:15:51.4069282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 20%] 2024-08-06T21:15:51.4070548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0076s] [ 20%] 2024-08-06T21:15:51.4071835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 20%] 2024-08-06T21:15:51.4073111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 20%] 2024-08-06T21:15:51.4074413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 20%] 2024-08-06T21:15:51.4075701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 20%] 2024-08-06T21:15:51.4077015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 20%] 2024-08-06T21:15:51.4078349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0062s] [ 20%] 2024-08-06T21:15:51.4079615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 20%] 2024-08-06T21:15:51.4080897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0072s] [ 20%] 2024-08-06T21:15:51.4082208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 20%] 2024-08-06T21:15:51.4083551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 20%] 2024-08-06T21:15:51.4084832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 20%] 2024-08-06T21:15:51.4086119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0064s] [ 20%] 2024-08-06T21:15:51.4087396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 20%] 2024-08-06T21:15:51.4088660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0062s] [ 20%] 2024-08-06T21:15:51.4089947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 21%] 2024-08-06T21:15:51.4091213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 21%] 2024-08-06T21:15:51.4092509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 21%] 2024-08-06T21:15:51.4093781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 21%] 2024-08-06T21:15:51.4095074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 21%] 2024-08-06T21:15:51.4096393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0062s] [ 21%] 2024-08-06T21:15:51.4097738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 21%] 2024-08-06T21:15:51.4098999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 21%] 2024-08-06T21:15:51.4100274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 21%] 2024-08-06T21:15:51.4101604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0076s] [ 21%] 2024-08-06T21:15:51.4102959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 21%] 2024-08-06T21:15:51.4104636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0076s] [ 21%] 2024-08-06T21:15:51.4105944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 21%] 2024-08-06T21:15:51.4107235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 21%] 2024-08-06T21:15:51.4108493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 21%] 2024-08-06T21:15:51.4109768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 21%] 2024-08-06T21:15:51.4111035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 21%] 2024-08-06T21:15:51.4112321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0071s] [ 21%] 2024-08-06T21:15:51.4113589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 21%] 2024-08-06T21:15:51.4114955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 21%] 2024-08-06T21:15:51.4116317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 21%] 2024-08-06T21:15:51.4117589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0211s] [ 21%] 2024-08-06T21:15:51.4118879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0231s] [ 21%] 2024-08-06T21:15:51.4120201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0197s] [ 21%] 2024-08-06T21:15:51.4121560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0197s] [ 21%] 2024-08-06T21:15:51.4122832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0312s] [ 21%] 2024-08-06T21:15:51.4124128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0321s] [ 21%] 2024-08-06T21:15:51.4125421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0227s] [ 21%] 2024-08-06T21:15:51.4126703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0228s] [ 21%] 2024-08-06T21:15:51.4127990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0228s] [ 21%] 2024-08-06T21:15:51.4129268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.1959s] [ 21%] 2024-08-06T21:15:51.4130564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0220s] [ 21%] 2024-08-06T21:15:51.4131840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0219s] [ 21%] 2024-08-06T21:15:51.4133127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0323s] [ 21%] 2024-08-06T21:15:51.4134477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0332s] [ 21%] 2024-08-06T21:15:51.4135821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0254s] [ 21%] 2024-08-06T21:15:51.4137098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0254s] [ 21%] 2024-08-06T21:15:51.4138431Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0256s] [ 21%] 2024-08-06T21:15:51.4139761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0276s] [ 21%] 2024-08-06T21:15:51.4141033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0265s] [ 21%] 2024-08-06T21:15:51.4142325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0264s] [ 21%] 2024-08-06T21:15:51.4143604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0352s] [ 21%] 2024-08-06T21:15:51.4144930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0361s] [ 21%] 2024-08-06T21:15:51.4146201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0301s] [ 21%] 2024-08-06T21:15:51.4147499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0306s] [ 21%] 2024-08-06T21:15:51.4148775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0207s] [ 21%] 2024-08-06T21:15:51.4150071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0221s] [ 21%] 2024-08-06T21:15:51.4151336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0192s] [ 21%] 2024-08-06T21:15:51.4152656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0189s] [ 21%] 2024-08-06T21:15:51.4153999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0307s] [ 21%] 2024-08-06T21:15:51.4155290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0313s] [ 21%] 2024-08-06T21:15:51.4156577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.1886s] [ 21%] 2024-08-06T21:15:51.4157896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0214s] [ 21%] 2024-08-06T21:15:51.4159248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 21%] 2024-08-06T21:15:51.4160516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 21%] 2024-08-06T21:15:51.4161803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 21%] 2024-08-06T21:15:51.4163079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 21%] 2024-08-06T21:15:51.4164378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0081s] [ 21%] 2024-08-06T21:15:51.4165651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 21%] 2024-08-06T21:15:51.4166919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0079s] [ 21%] 2024-08-06T21:15:51.4168217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 21%] 2024-08-06T21:15:51.4169478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0067s] [ 21%] 2024-08-06T21:15:51.4170808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0070s] [ 21%] 2024-08-06T21:15:51.4172125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0077s] [ 21%] 2024-08-06T21:15:51.4173415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 21%] 2024-08-06T21:15:51.4174706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0086s] [ 21%] 2024-08-06T21:15:51.4176043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 21%] 2024-08-06T21:15:51.4177363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0083s] [ 21%] 2024-08-06T21:15:51.4178638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0086s] [ 21%] 2024-08-06T21:15:51.4179919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0073s] [ 21%] 2024-08-06T21:15:51.4181190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 21%] 2024-08-06T21:15:51.4182480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0084s] [ 21%] 2024-08-06T21:15:51.4183747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0081s] [ 21%] 2024-08-06T21:15:51.4185052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0097s] [ 21%] 2024-08-06T21:15:51.4186331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0097s] [ 21%] 2024-08-06T21:15:51.4187618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0094s] [ 21%] 2024-08-06T21:15:51.4188885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0089s] [ 21%] 2024-08-06T21:15:51.4190211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 22%] 2024-08-06T21:15:51.4191532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 22%] 2024-08-06T21:15:51.4192790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0069s] [ 22%] 2024-08-06T21:15:51.4194083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 22%] 2024-08-06T21:15:51.4195394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0080s] [ 22%] 2024-08-06T21:15:51.4196739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0081s] [ 22%] 2024-08-06T21:15:51.4198000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 22%] 2024-08-06T21:15:51.4199291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 22%] 2024-08-06T21:15:51.4200559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 22%] 2024-08-06T21:15:51.4201845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 22%] 2024-08-06T21:15:51.4203141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 22%] 2024-08-06T21:15:51.4204797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 22%] 2024-08-06T21:15:51.4206113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 22%] 2024-08-06T21:15:51.4207377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 22%] 2024-08-06T21:15:51.4208732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 22%] 2024-08-06T21:15:51.4210074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 22%] 2024-08-06T21:15:51.4211354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 22%] 2024-08-06T21:15:51.4212618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 22%] 2024-08-06T21:15:51.4213947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 22%] 2024-08-06T21:15:51.4215302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 22%] 2024-08-06T21:15:51.4216584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 22%] 2024-08-06T21:15:51.4217849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0070s] [ 22%] 2024-08-06T21:15:51.4219113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 22%] 2024-08-06T21:15:51.4220409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 22%] 2024-08-06T21:15:51.4221668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 22%] 2024-08-06T21:15:51.4222953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 22%] 2024-08-06T21:15:51.4224223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 22%] 2024-08-06T21:15:51.4225524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 22%] 2024-08-06T21:15:51.4226791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0077s] [ 22%] 2024-08-06T21:15:51.4228122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 22%] 2024-08-06T21:15:51.4229444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 22%] 2024-08-06T21:15:51.4230713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 22%] 2024-08-06T21:15:51.4231990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 22%] 2024-08-06T21:15:51.4233311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 22%] 2024-08-06T21:15:51.4234648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 22%] 2024-08-06T21:15:51.4235912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 22%] 2024-08-06T21:15:51.4237197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0071s] [ 22%] 2024-08-06T21:15:51.4238473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0070s] [ 22%] 2024-08-06T21:15:51.4239757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 22%] 2024-08-06T21:15:51.4241020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 22%] 2024-08-06T21:15:51.4242320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0089s] [ 22%] 2024-08-06T21:15:51.4243602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0092s] [ 22%] 2024-08-06T21:15:51.4244894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0089s] [ 22%] 2024-08-06T21:15:51.4246231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0089s] [ 22%] 2024-08-06T21:15:51.4247556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0119s] [ 22%] 2024-08-06T21:15:51.4248854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0119s] [ 22%] 2024-08-06T21:15:51.4250124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0101s] [ 22%] 2024-08-06T21:15:51.4251465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0101s] [ 22%] 2024-08-06T21:15:51.4252786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0092s] [ 22%] 2024-08-06T21:15:51.4254077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0097s] [ 22%] 2024-08-06T21:15:51.4255364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0097s] [ 22%] 2024-08-06T21:15:51.4256646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0097s] [ 22%] 2024-08-06T21:15:51.4257944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0124s] [ 22%] 2024-08-06T21:15:51.4259218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0125s] [ 22%] 2024-08-06T21:15:51.4260515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0114s] [ 22%] 2024-08-06T21:15:51.4261798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0117s] [ 22%] 2024-08-06T21:15:51.4263084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0108s] [ 22%] 2024-08-06T21:15:51.4264361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0119s] [ 22%] 2024-08-06T21:15:51.4265698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0125s] [ 22%] 2024-08-06T21:15:51.4267029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0123s] [ 22%] 2024-08-06T21:15:51.4268318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0135s] [ 22%] 2024-08-06T21:15:51.4269599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0136s] [ 22%] 2024-08-06T21:15:51.4270985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0135s] [ 22%] 2024-08-06T21:15:51.4272336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0130s] [ 22%] 2024-08-06T21:15:51.4273608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0086s] [ 22%] 2024-08-06T21:15:51.4274932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0092s] [ 22%] 2024-08-06T21:15:51.4276209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0088s] [ 22%] 2024-08-06T21:15:51.4277497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0088s] [ 22%] 2024-08-06T21:15:51.4278768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0121s] [ 22%] 2024-08-06T21:15:51.4280066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0123s] [ 22%] 2024-08-06T21:15:51.4281345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0108s] [ 22%] 2024-08-06T21:15:51.4282620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0107s] [ 22%] 2024-08-06T21:15:51.4283954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0146s] [ 22%] 2024-08-06T21:15:51.4285303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0156s] [ 22%] 2024-08-06T21:15:51.4286588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0154s] [ 22%] 2024-08-06T21:15:51.4287860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0144s] [ 22%] 2024-08-06T21:15:51.4289193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0200s] [ 22%] 2024-08-06T21:15:51.4290524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0205s] [ 22%] 2024-08-06T21:15:51.4291811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0161s] [ 23%] 2024-08-06T21:15:51.4293090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0161s] [ 23%] 2024-08-06T21:15:51.4294407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0149s] [ 23%] 2024-08-06T21:15:51.4295691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0163s] [ 23%] 2024-08-06T21:15:51.4296956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0160s] [ 23%] 2024-08-06T21:15:51.4298251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0158s] [ 23%] 2024-08-06T21:15:51.4299534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0208s] [ 23%] 2024-08-06T21:15:51.4300831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0210s] [ 23%] 2024-08-06T21:15:51.4302097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0177s] [ 23%] 2024-08-06T21:15:51.4303479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0177s] [ 23%] 2024-08-06T21:15:51.4305202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0174s] [ 23%] 2024-08-06T21:15:51.4306525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0186s] [ 23%] 2024-08-06T21:15:51.4307917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0184s] [ 23%] 2024-08-06T21:15:51.4309285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0184s] [ 23%] 2024-08-06T21:15:51.4310556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0228s] [ 23%] 2024-08-06T21:15:51.4311829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0230s] [ 23%] 2024-08-06T21:15:51.4313125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0204s] [ 23%] 2024-08-06T21:15:51.4314409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0204s] [ 23%] 2024-08-06T21:15:51.4315713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0134s] [ 23%] 2024-08-06T21:15:51.4316984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0139s] [ 23%] 2024-08-06T21:15:51.4318273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0136s] [ 23%] 2024-08-06T21:15:51.4319547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0136s] [ 23%] 2024-08-06T21:15:51.4320835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0194s] [ 23%] 2024-08-06T21:15:51.4322172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0199s] [ 23%] 2024-08-06T21:15:51.4323517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0153s] [ 23%] 2024-08-06T21:15:51.4324815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0153s] [ 23%] 2024-08-06T21:15:51.4326080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 23%] 2024-08-06T21:15:51.4327407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 23%] 2024-08-06T21:15:51.4328720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0062s] [ 23%] 2024-08-06T21:15:51.4330003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 23%] 2024-08-06T21:15:51.4331267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0079s] [ 23%] 2024-08-06T21:15:51.4332562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 23%] 2024-08-06T21:15:51.4333835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 23%] 2024-08-06T21:15:51.4335140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 23%] 2024-08-06T21:15:51.4336412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 23%] 2024-08-06T21:15:51.4337689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 23%] 2024-08-06T21:15:51.4338963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0064s] [ 23%] 2024-08-06T21:15:51.4340274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 23%] 2024-08-06T21:15:51.4341619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0081s] [ 23%] 2024-08-06T21:15:51.4342902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 23%] 2024-08-06T21:15:51.4344185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 23%] 2024-08-06T21:15:51.4345522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 23%] 2024-08-06T21:15:51.4346860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0071s] [ 23%] 2024-08-06T21:15:51.4348129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 23%] 2024-08-06T21:15:51.4349385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0075s] [ 23%] 2024-08-06T21:15:51.4350672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 23%] 2024-08-06T21:15:51.4351948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0088s] [ 23%] 2024-08-06T21:15:51.4353234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0088s] [ 23%] 2024-08-06T21:15:51.4354501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0080s] [ 23%] 2024-08-06T21:15:51.4355808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0081s] [ 23%] 2024-08-06T21:15:51.4357075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 23%] 2024-08-06T21:15:51.4358352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 23%] 2024-08-06T21:15:51.4359657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 23%] 2024-08-06T21:15:51.4360998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 23%] 2024-08-06T21:15:51.4362263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0079s] [ 23%] 2024-08-06T21:15:51.4363533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0078s] [ 23%] 2024-08-06T21:15:51.4364871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 23%] 2024-08-06T21:15:51.4366196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 23%] 2024-08-06T21:15:51.4367481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 23%] 2024-08-06T21:15:51.4368753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 23%] 2024-08-06T21:15:51.4370036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 23%] 2024-08-06T21:15:51.4371301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 23%] 2024-08-06T21:15:51.4372576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 23%] 2024-08-06T21:15:51.4373849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 23%] 2024-08-06T21:15:51.4375142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 23%] 2024-08-06T21:15:51.4376430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 23%] 2024-08-06T21:15:51.4377743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 23%] 2024-08-06T21:15:51.4379090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 23%] 2024-08-06T21:15:51.4380359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 23%] 2024-08-06T21:15:51.4381644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 23%] 2024-08-06T21:15:51.4382950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0076s] [ 23%] 2024-08-06T21:15:51.4384288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 23%] 2024-08-06T21:15:51.4385570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 23%] 2024-08-06T21:15:51.4386850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 23%] 2024-08-06T21:15:51.4388117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0066s] [ 23%] 2024-08-06T21:15:51.4389391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 23%] 2024-08-06T21:15:51.4390670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0066s] [ 23%] 2024-08-06T21:15:51.4391940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 24%] 2024-08-06T21:15:51.4393230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0082s] [ 24%] 2024-08-06T21:15:51.4394501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 24%] 2024-08-06T21:15:51.4395781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0073s] [ 24%] 2024-08-06T21:15:51.4397092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 24%] 2024-08-06T21:15:51.4398425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 24%] 2024-08-06T21:15:51.4399695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 24%] 2024-08-06T21:15:51.4400957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 24%] 2024-08-06T21:15:51.4402316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 24%] 2024-08-06T21:15:51.4403641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0073s] [ 24%] 2024-08-06T21:15:51.4405301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 24%] 2024-08-06T21:15:51.4406595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 24%] 2024-08-06T21:15:51.4407884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 24%] 2024-08-06T21:15:51.4409152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0251s] [ 24%] 2024-08-06T21:15:51.4410442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.1982s] [ 24%] 2024-08-06T21:15:51.4411714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0252s] [ 24%] 2024-08-06T21:15:51.4413010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0252s] [ 24%] 2024-08-06T21:15:51.4414281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0377s] [ 24%] 2024-08-06T21:15:51.4415660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0392s] [ 24%] 2024-08-06T21:15:51.4417019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0289s] [ 24%] 2024-08-06T21:15:51.4418301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0290s] [ 24%] 2024-08-06T21:15:51.4419592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0287s] [ 24%] 2024-08-06T21:15:51.4420933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0312s] [ 24%] 2024-08-06T21:15:51.4422296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0298s] [ 24%] 2024-08-06T21:15:51.4423569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0301s] [ 24%] 2024-08-06T21:15:51.4424856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0404s] [ 24%] 2024-08-06T21:15:51.4426146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0415s] [ 24%] 2024-08-06T21:15:51.4427426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0322s] [ 24%] 2024-08-06T21:15:51.4428860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0329s] [ 24%] 2024-08-06T21:15:51.4430146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0352s] [ 24%] 2024-08-06T21:15:51.4431448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0356s] [ 24%] 2024-08-06T21:15:51.4432718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0360s] [ 24%] 2024-08-06T21:15:51.4434008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0346s] [ 24%] 2024-08-06T21:15:51.4435345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.2120s] [ 24%] 2024-08-06T21:15:51.4436696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0476s] [ 24%] 2024-08-06T21:15:51.4437965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0395s] [ 24%] 2024-08-06T21:15:51.4439263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0388s] [ 24%] 2024-08-06T21:15:51.4440572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0252s] [ 24%] 2024-08-06T21:15:51.4441901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0267s] [ 24%] 2024-08-06T21:15:51.4443185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0253s] [ 24%] 2024-08-06T21:15:51.4444458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0255s] [ 24%] 2024-08-06T21:15:51.4445779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0371s] [ 24%] 2024-08-06T21:15:51.4447058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0381s] [ 24%] 2024-08-06T21:15:51.4448343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0282s] [ 24%] 2024-08-06T21:15:51.4449616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0290s] [ 24%] 2024-08-06T21:15:51.4450905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0077s] [ 24%] 2024-08-06T21:15:51.4452172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 24%] 2024-08-06T21:15:51.4453479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0081s] [ 24%] 2024-08-06T21:15:51.4454811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 24%] 2024-08-06T21:15:51.4456094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0096s] [ 24%] 2024-08-06T21:15:51.4457382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0089s] [ 24%] 2024-08-06T21:15:51.4458691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0086s] [ 24%] 2024-08-06T21:15:51.4460660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0085s] [ 24%] 2024-08-06T21:15:51.4461925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0081s] [ 24%] 2024-08-06T21:15:51.4463211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 24%] 2024-08-06T21:15:51.4464484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0081s] [ 24%] 2024-08-06T21:15:51.4465788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 24%] 2024-08-06T21:15:51.4467063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0099s] [ 24%] 2024-08-06T21:15:51.4468337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0099s] [ 24%] 2024-08-06T21:15:51.4469624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0088s] [ 24%] 2024-08-06T21:15:51.4470893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0087s] [ 24%] 2024-08-06T21:15:51.4472175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0085s] [ 24%] 2024-08-06T21:15:51.4473492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0085s] [ 24%] 2024-08-06T21:15:51.4474827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0087s] [ 24%] 2024-08-06T21:15:51.4476118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0087s] [ 24%] 2024-08-06T21:15:51.4477404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0091s] [ 24%] 2024-08-06T21:15:51.4478721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0091s] [ 24%] 2024-08-06T21:15:51.4480036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0090s] [ 24%] 2024-08-06T21:15:51.4481324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0091s] [ 24%] 2024-08-06T21:15:51.4482590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 24%] 2024-08-06T21:15:51.4483889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 24%] 2024-08-06T21:15:51.4485157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0064s] [ 24%] 2024-08-06T21:15:51.4486460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 24%] 2024-08-06T21:15:51.4487729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0079s] [ 24%] 2024-08-06T21:15:51.4489019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 24%] 2024-08-06T21:15:51.4490283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0072s] [ 24%] 2024-08-06T21:15:51.4491619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 24%] 2024-08-06T21:15:51.4492936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 25%] 2024-08-06T21:15:51.4494203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 25%] 2024-08-06T21:15:51.4495481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 25%] 2024-08-06T21:15:51.4496795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 25%] 2024-08-06T21:15:51.4498132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 25%] 2024-08-06T21:15:51.4499402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 25%] 2024-08-06T21:15:51.4500685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 25%] 2024-08-06T21:15:51.4501966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 25%] 2024-08-06T21:15:51.4503285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 25%] 2024-08-06T21:15:51.4504897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 25%] 2024-08-06T21:15:51.4506214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0062s] [ 25%] 2024-08-06T21:15:51.4507511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 25%] 2024-08-06T21:15:51.4508776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0082s] [ 25%] 2024-08-06T21:15:51.4510066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0088s] [ 25%] 2024-08-06T21:15:51.4511406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0076s] [ 25%] 2024-08-06T21:15:51.4512766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 25%] 2024-08-06T21:15:51.4514031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0071s] [ 25%] 2024-08-06T21:15:51.4515316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 25%] 2024-08-06T21:15:51.4516673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0069s] [ 25%] 2024-08-06T21:15:51.4518025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 25%] 2024-08-06T21:15:51.4519294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0089s] [ 25%] 2024-08-06T21:15:51.4520576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0089s] [ 25%] 2024-08-06T21:15:51.4521863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0078s] [ 25%] 2024-08-06T21:15:51.4523133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 25%] 2024-08-06T21:15:51.4524407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 25%] 2024-08-06T21:15:51.4525676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 25%] 2024-08-06T21:15:51.4526972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 25%] 2024-08-06T21:15:51.4528236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 25%] 2024-08-06T21:15:51.4529563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0079s] [ 25%] 2024-08-06T21:15:51.4530888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 25%] 2024-08-06T21:15:51.4532154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 25%] 2024-08-06T21:15:51.4533435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_152_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 25%] 2024-08-06T21:15:51.4534750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0121s] [ 25%] 2024-08-06T21:15:51.4536115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0129s] [ 25%] 2024-08-06T21:15:51.4537387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0119s] [ 25%] 2024-08-06T21:15:51.4538680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0121s] [ 25%] 2024-08-06T21:15:51.4539958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0173s] [ 25%] 2024-08-06T21:15:51.4541259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0176s] [ 25%] 2024-08-06T21:15:51.4542529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0135s] [ 25%] 2024-08-06T21:15:51.4543828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0137s] [ 25%] 2024-08-06T21:15:51.4545105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0125s] [ 25%] 2024-08-06T21:15:51.4546403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0128s] [ 25%] 2024-08-06T21:15:51.4547692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0126s] [ 25%] 2024-08-06T21:15:51.4549011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0129s] [ 25%] 2024-08-06T21:15:51.4550358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.1891s] [ 25%] 2024-08-06T21:15:51.4551639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0183s] [ 25%] 2024-08-06T21:15:51.4552971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0155s] [ 25%] 2024-08-06T21:15:51.4554300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0156s] [ 25%] 2024-08-06T21:15:51.4555589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0145s] [ 25%] 2024-08-06T21:15:51.4556878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0162s] [ 25%] 2024-08-06T21:15:51.4558150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0172s] [ 25%] 2024-08-06T21:15:51.4559454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0160s] [ 25%] 2024-08-06T21:15:51.4560724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0198s] [ 25%] 2024-08-06T21:15:51.4562016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0198s] [ 25%] 2024-08-06T21:15:51.4563289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0177s] [ 25%] 2024-08-06T21:15:51.4564587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0174s] [ 25%] 2024-08-06T21:15:51.4565877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0117s] [ 25%] 2024-08-06T21:15:51.4567214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0122s] [ 25%] 2024-08-06T21:15:51.4568598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0120s] [ 25%] 2024-08-06T21:15:51.4569889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0120s] [ 25%] 2024-08-06T21:15:51.4571157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0169s] [ 25%] 2024-08-06T21:15:51.4572474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0171s] [ 25%] 2024-08-06T21:15:51.4573823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0138s] [ 25%] 2024-08-06T21:15:51.4575104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0134s] [ 25%] 2024-08-06T21:15:51.4576420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0203s] [ 25%] 2024-08-06T21:15:51.4577708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0216s] [ 25%] 2024-08-06T21:15:51.4578996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0199s] [ 25%] 2024-08-06T21:15:51.4580270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0198s] [ 25%] 2024-08-06T21:15:51.4581566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0301s] [ 25%] 2024-08-06T21:15:51.4582853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0311s] [ 25%] 2024-08-06T21:15:51.4584126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0232s] [ 25%] 2024-08-06T21:15:51.4585469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0235s] [ 25%] 2024-08-06T21:15:51.4586826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0208s] [ 25%] 2024-08-06T21:15:51.4588128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0229s] [ 25%] 2024-08-06T21:15:51.4589399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0216s] [ 25%] 2024-08-06T21:15:51.4590732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0213s] [ 25%] 2024-08-06T21:15:51.4592059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.1989s] [ 25%] 2024-08-06T21:15:51.4593357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0305s] [ 25%] 2024-08-06T21:15:51.4594631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0237s] [ 26%] 2024-08-06T21:15:51.4595942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0238s] [ 26%] 2024-08-06T21:15:51.4597229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0231s] [ 26%] 2024-08-06T21:15:51.4598507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0253s] [ 26%] 2024-08-06T21:15:51.4599797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0246s] [ 26%] 2024-08-06T21:15:51.4601077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0248s] [ 26%] 2024-08-06T21:15:51.4602404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0330s] [ 26%] 2024-08-06T21:15:51.4603694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0334s] [ 26%] 2024-08-06T21:15:51.4605432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0284s] [ 26%] 2024-08-06T21:15:51.4606858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0282s] [ 26%] 2024-08-06T21:15:51.4608155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0186s] [ 26%] 2024-08-06T21:15:51.4609433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0194s] [ 26%] 2024-08-06T21:15:51.4610784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0177s] [ 26%] 2024-08-06T21:15:51.4612117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0177s] [ 26%] 2024-08-06T21:15:51.4613389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0283s] [ 26%] 2024-08-06T21:15:51.4614687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0290s] [ 26%] 2024-08-06T21:15:51.4615965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0211s] [ 26%] 2024-08-06T21:15:51.4617273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0211s] [ 26%] 2024-08-06T21:15:51.4618538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0081s] [ 26%] 2024-08-06T21:15:51.4619829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 26%] 2024-08-06T21:15:51.4621096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0080s] [ 26%] 2024-08-06T21:15:51.4622383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 26%] 2024-08-06T21:15:51.4623696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0108s] [ 26%] 2024-08-06T21:15:51.4625027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0112s] [ 26%] 2024-08-06T21:15:51.4626321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0093s] [ 26%] 2024-08-06T21:15:51.4627616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0092s] [ 26%] 2024-08-06T21:15:51.4628944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0087s] [ 26%] 2024-08-06T21:15:51.4630269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.1815s] [ 26%] 2024-08-06T21:15:51.4631555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0081s] [ 26%] 2024-08-06T21:15:51.4632821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 26%] 2024-08-06T21:15:51.4634106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0110s] [ 26%] 2024-08-06T21:15:51.4635383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0110s] [ 26%] 2024-08-06T21:15:51.4636687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0091s] [ 26%] 2024-08-06T21:15:51.4637964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0089s] [ 26%] 2024-08-06T21:15:51.4639234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0091s] [ 26%] 2024-08-06T21:15:51.4640520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0091s] [ 26%] 2024-08-06T21:15:51.4641781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0095s] [ 26%] 2024-08-06T21:15:51.4643114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0095s] [ 26%] 2024-08-06T21:15:51.4644434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0118s] [ 26%] 2024-08-06T21:15:51.4645728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0120s] [ 26%] 2024-08-06T21:15:51.4647022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0101s] [ 26%] 2024-08-06T21:15:51.4648362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0101s] [ 26%] 2024-08-06T21:15:51.4649674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0072s] [ 26%] 2024-08-06T21:15:51.4650944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 26%] 2024-08-06T21:15:51.4652230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0072s] [ 26%] 2024-08-06T21:15:51.4653503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 26%] 2024-08-06T21:15:51.4654789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0101s] [ 26%] 2024-08-06T21:15:51.4656054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0101s] [ 26%] 2024-08-06T21:15:51.4657353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0081s] [ 26%] 2024-08-06T21:15:51.4658627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 26%] 2024-08-06T21:15:51.4659906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0070s] [ 26%] 2024-08-06T21:15:51.4661233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0070s] [ 26%] 2024-08-06T21:15:51.4662570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0071s] [ 26%] 2024-08-06T21:15:51.4663838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 26%] 2024-08-06T21:15:51.4665099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0094s] [ 26%] 2024-08-06T21:15:51.4666457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0094s] [ 26%] 2024-08-06T21:15:51.4667788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0081s] [ 26%] 2024-08-06T21:15:51.4669074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 26%] 2024-08-06T21:15:51.4670949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0080s] [ 26%] 2024-08-06T21:15:51.4672288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 26%] 2024-08-06T21:15:51.4673562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 26%] 2024-08-06T21:15:51.4674880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 26%] 2024-08-06T21:15:51.4676158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0092s] [ 26%] 2024-08-06T21:15:51.4677481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0092s] [ 26%] 2024-08-06T21:15:51.4678747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0075s] [ 26%] 2024-08-06T21:15:51.4680030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 26%] 2024-08-06T21:15:51.4681366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0081s] [ 26%] 2024-08-06T21:15:51.4682694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0085s] [ 26%] 2024-08-06T21:15:51.4683970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0079s] [ 26%] 2024-08-06T21:15:51.4685237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 26%] 2024-08-06T21:15:51.4686567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0100s] [ 26%] 2024-08-06T21:15:51.4687907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0104s] [ 26%] 2024-08-06T21:15:51.4689188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0083s] [ 26%] 2024-08-06T21:15:51.4690457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0086s] [ 26%] 2024-08-06T21:15:51.4691742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 26%] 2024-08-06T21:15:51.4693007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 26%] 2024-08-06T21:15:51.4694278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0067s] [ 26%] 2024-08-06T21:15:51.4695544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 27%] 2024-08-06T21:15:51.4696811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0093s] [ 27%] 2024-08-06T21:15:51.4698104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0095s] [ 27%] 2024-08-06T21:15:51.4699421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0076s] [ 27%] 2024-08-06T21:15:51.4700773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 27%] 2024-08-06T21:15:51.4702043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0365s] [ 27%] 2024-08-06T21:15:51.4703377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0392s] [ 27%] 2024-08-06T21:15:51.4705124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0347s] [ 27%] 2024-08-06T21:15:51.4706523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0338s] [ 27%] 2024-08-06T21:15:51.4707805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0550s] [ 27%] 2024-08-06T21:15:51.4709090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0573s] [ 27%] 2024-08-06T21:15:51.4710375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0395s] [ 27%] 2024-08-06T21:15:51.4711653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0398s] [ 27%] 2024-08-06T21:15:51.4712939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0376s] [ 27%] 2024-08-06T21:15:51.4714219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0407s] [ 27%] 2024-08-06T21:15:51.4715511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.2028s] [ 27%] 2024-08-06T21:15:51.4716787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0377s] [ 27%] 2024-08-06T21:15:51.4718101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0563s] [ 27%] 2024-08-06T21:15:51.4719444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0585s] [ 27%] 2024-08-06T21:15:51.4720807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0439s] [ 27%] 2024-08-06T21:15:51.4722090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0444s] [ 27%] 2024-08-06T21:15:51.4723404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0434s] [ 27%] 2024-08-06T21:15:51.4724777Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0471s] [ 27%] 2024-08-06T21:15:51.4726049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0444s] [ 27%] 2024-08-06T21:15:51.4727342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0450s] [ 27%] 2024-08-06T21:15:51.4728635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0616s] [ 27%] 2024-08-06T21:15:51.4729938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.2312s] [ 27%] 2024-08-06T21:15:51.4731211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0511s] [ 27%] 2024-08-06T21:15:51.4732510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0514s] [ 27%] 2024-08-06T21:15:51.4733785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0343s] [ 27%] 2024-08-06T21:15:51.4735080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0360s] [ 27%] 2024-08-06T21:15:51.4736353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0317s] [ 27%] 2024-08-06T21:15:51.4737681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0316s] [ 27%] 2024-08-06T21:15:51.4739024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0542s] [ 27%] 2024-08-06T21:15:51.4740306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0556s] [ 27%] 2024-08-06T21:15:51.4741592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0371s] [ 27%] 2024-08-06T21:15:51.4742915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0376s] [ 27%] 2024-08-06T21:15:51.4744252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0082s] [ 27%] 2024-08-06T21:15:51.4745525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 27%] 2024-08-06T21:15:51.4746807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0081s] [ 27%] 2024-08-06T21:15:51.4748109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0081s] [ 27%] 2024-08-06T21:15:51.4749382Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0110s] [ 27%] 2024-08-06T21:15:51.4750674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0108s] [ 27%] 2024-08-06T21:15:51.4751946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0091s] [ 27%] 2024-08-06T21:15:51.4753244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0091s] [ 27%] 2024-08-06T21:15:51.4754507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0086s] [ 27%] 2024-08-06T21:15:51.4755838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.1773s] [ 27%] 2024-08-06T21:15:51.4757157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0091s] [ 27%] 2024-08-06T21:15:51.4758464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0090s] [ 27%] 2024-08-06T21:15:51.4759731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0111s] [ 27%] 2024-08-06T21:15:51.4761062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0111s] [ 27%] 2024-08-06T21:15:51.4762380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0100s] [ 27%] 2024-08-06T21:15:51.4763656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0098s] [ 27%] 2024-08-06T21:15:51.4764933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0091s] [ 27%] 2024-08-06T21:15:51.4766205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0095s] [ 27%] 2024-08-06T21:15:51.4767519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0109s] [ 27%] 2024-08-06T21:15:51.4768791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0104s] [ 27%] 2024-08-06T21:15:51.4770078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0120s] [ 27%] 2024-08-06T21:15:51.4771355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0121s] [ 27%] 2024-08-06T21:15:51.4772646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0113s] [ 27%] 2024-08-06T21:15:51.4773917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0114s] [ 27%] 2024-08-06T21:15:51.4775222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0075s] [ 27%] 2024-08-06T21:15:51.4776562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 27%] 2024-08-06T21:15:51.4777848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0078s] [ 27%] 2024-08-06T21:15:51.4779133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0077s] [ 27%] 2024-08-06T21:15:51.4780440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0103s] [ 27%] 2024-08-06T21:15:51.4781780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0112s] [ 27%] 2024-08-06T21:15:51.4783043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0087s] [ 27%] 2024-08-06T21:15:51.4784342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0087s] [ 27%] 2024-08-06T21:15:51.4785615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0070s] [ 27%] 2024-08-06T21:15:51.4786899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0070s] [ 27%] 2024-08-06T21:15:51.4788181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0076s] [ 27%] 2024-08-06T21:15:51.4789451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 27%] 2024-08-06T21:15:51.4790736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0104s] [ 27%] 2024-08-06T21:15:51.4792000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0099s] [ 27%] 2024-08-06T21:15:51.4793321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0081s] [ 27%] 2024-08-06T21:15:51.4794642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 27%] 2024-08-06T21:15:51.4795930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0077s] [ 27%] 2024-08-06T21:15:51.4797197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 28%] 2024-08-06T21:15:51.4798543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0066s] [ 28%] 2024-08-06T21:15:51.4799862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 28%] 2024-08-06T21:15:51.4801125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0094s] [ 28%] 2024-08-06T21:15:51.4802447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0093s] [ 28%] 2024-08-06T21:15:51.4803724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0073s] [ 28%] 2024-08-06T21:15:51.4805396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 28%] 2024-08-06T21:15:51.4806673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0081s] [ 28%] 2024-08-06T21:15:51.4807973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 28%] 2024-08-06T21:15:51.4809242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0079s] [ 28%] 2024-08-06T21:15:51.4810533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 28%] 2024-08-06T21:15:51.4811792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0101s] [ 28%] 2024-08-06T21:15:51.4813157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0101s] [ 28%] 2024-08-06T21:15:51.4814496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0085s] [ 28%] 2024-08-06T21:15:51.4815766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 28%] 2024-08-06T21:15:51.4817043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 28%] 2024-08-06T21:15:51.4818387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 28%] 2024-08-06T21:15:51.4819738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 28%] 2024-08-06T21:15:51.4821004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 28%] 2024-08-06T21:15:51.4822288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0087s] [ 28%] 2024-08-06T21:15:51.4823563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0087s] [ 28%] 2024-08-06T21:15:51.4824840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 28%] 2024-08-06T21:15:51.4826105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 28%] 2024-08-06T21:15:51.4827377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 28%] 2024-08-06T21:15:51.4828685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 28%] 2024-08-06T21:15:51.4829945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 28%] 2024-08-06T21:15:51.4831282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 28%] 2024-08-06T21:15:51.4832600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 28%] 2024-08-06T21:15:51.4833889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 28%] 2024-08-06T21:15:51.4835151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 28%] 2024-08-06T21:15:51.4836479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 28%] 2024-08-06T21:15:51.4837791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 28%] 2024-08-06T21:15:51.4839086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 28%] 2024-08-06T21:15:51.4840340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 28%] 2024-08-06T21:15:51.4841606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 28%] 2024-08-06T21:15:51.4842895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 28%] 2024-08-06T21:15:51.4844162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 28%] 2024-08-06T21:15:51.4845443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 28%] 2024-08-06T21:15:51.4846719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 28%] 2024-08-06T21:15:51.4848013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 28%] 2024-08-06T21:15:51.4849281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 28%] 2024-08-06T21:15:51.4850600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 28%] 2024-08-06T21:15:51.4851918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 28%] 2024-08-06T21:15:51.4853182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 28%] 2024-08-06T21:15:51.4854468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 28%] 2024-08-06T21:15:51.4855770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 28%] 2024-08-06T21:15:51.4857104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 28%] 2024-08-06T21:15:51.4858387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 28%] 2024-08-06T21:15:51.4859669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 28%] 2024-08-06T21:15:51.4860929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 28%] 2024-08-06T21:15:51.4862200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 28%] 2024-08-06T21:15:51.4863454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 28%] 2024-08-06T21:15:51.4864725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 28%] 2024-08-06T21:15:51.4866010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 28%] 2024-08-06T21:15:51.4867271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 28%] 2024-08-06T21:15:51.4868561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0064s] [ 28%] 2024-08-06T21:15:51.4869937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 28%] 2024-08-06T21:15:51.4871277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 28%] 2024-08-06T21:15:51.4872540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 28%] 2024-08-06T21:15:51.4873861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 28%] 2024-08-06T21:15:51.4875180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 28%] 2024-08-06T21:15:51.4876451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 28%] 2024-08-06T21:15:51.4877722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 28%] 2024-08-06T21:15:51.4878988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0069s] [ 28%] 2024-08-06T21:15:51.4880278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 28%] 2024-08-06T21:15:51.4881535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 28%] 2024-08-06T21:15:51.4882823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 28%] 2024-08-06T21:15:51.4884091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0079s] [ 28%] 2024-08-06T21:15:51.4885380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 28%] 2024-08-06T21:15:51.4886639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 28%] 2024-08-06T21:15:51.4887982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 28%] 2024-08-06T21:15:51.4889293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0078s] [ 28%] 2024-08-06T21:15:51.4890559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 28%] 2024-08-06T21:15:51.4891830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 28%] 2024-08-06T21:15:51.4893137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 28%] 2024-08-06T21:15:51.4894468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0085s] [ 28%] 2024-08-06T21:15:51.4895735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0085s] [ 28%] 2024-08-06T21:15:51.4897015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0072s] [ 29%] 2024-08-06T21:15:51.4898310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 29%] 2024-08-06T21:15:51.4899592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 29%] 2024-08-06T21:15:51.4900851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 29%] 2024-08-06T21:15:51.4902130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 29%] 2024-08-06T21:15:51.4903453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 29%] 2024-08-06T21:15:51.4905101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0068s] [ 29%] 2024-08-06T21:15:51.4906395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 29%] 2024-08-06T21:15:51.4907861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 29%] 2024-08-06T21:15:51.4909229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 29%] 2024-08-06T21:15:51.4910503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 29%] 2024-08-06T21:15:51.4911819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 29%] 2024-08-06T21:15:51.4913145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 29%] 2024-08-06T21:15:51.4914421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 29%] 2024-08-06T21:15:51.4915682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 29%] 2024-08-06T21:15:51.4916966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 29%] 2024-08-06T21:15:51.4918251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 29%] 2024-08-06T21:15:51.4919537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 29%] 2024-08-06T21:15:51.4920795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 29%] 2024-08-06T21:15:51.4922078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 29%] 2024-08-06T21:15:51.4923338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 29%] 2024-08-06T21:15:51.4924595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 29%] 2024-08-06T21:15:51.4925918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 29%] 2024-08-06T21:15:51.4927239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 29%] 2024-08-06T21:15:51.4928530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 29%] 2024-08-06T21:15:51.4929793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 29%] 2024-08-06T21:15:51.4931106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 29%] 2024-08-06T21:15:51.4932417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 29%] 2024-08-06T21:15:51.4933685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 29%] 2024-08-06T21:15:51.4934949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 29%] 2024-08-06T21:15:51.4936235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 29%] 2024-08-06T21:15:51.4937505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 29%] 2024-08-06T21:15:51.4938786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 29%] 2024-08-06T21:15:51.4940070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 29%] 2024-08-06T21:15:51.4941327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 29%] 2024-08-06T21:15:51.4942596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 29%] 2024-08-06T21:15:51.4943844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 29%] 2024-08-06T21:15:51.4945202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 29%] 2024-08-06T21:15:51.4946517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 29%] 2024-08-06T21:15:51.4947798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 29%] 2024-08-06T21:15:51.4949077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 29%] 2024-08-06T21:15:51.4950383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 29%] 2024-08-06T21:15:51.4951698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 29%] 2024-08-06T21:15:51.4952954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 29%] 2024-08-06T21:15:51.4954227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 29%] 2024-08-06T21:15:51.4955492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 29%] 2024-08-06T21:15:51.4956788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 29%] 2024-08-06T21:15:51.4958048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 29%] 2024-08-06T21:15:51.4959335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 29%] 2024-08-06T21:15:51.4960600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 29%] 2024-08-06T21:15:51.4961855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 29%] 2024-08-06T21:15:51.4963176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 29%] 2024-08-06T21:15:51.4964477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 29%] 2024-08-06T21:15:51.4965748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 29%] 2024-08-06T21:15:51.4966999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 29%] 2024-08-06T21:15:51.4968330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 29%] 2024-08-06T21:15:51.4969637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 29%] 2024-08-06T21:15:51.4970917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 29%] 2024-08-06T21:15:51.4972164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 29%] 2024-08-06T21:15:51.4973445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 29%] 2024-08-06T21:15:51.4974700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 29%] 2024-08-06T21:15:51.4975953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 29%] 2024-08-06T21:15:51.4977230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 29%] 2024-08-06T21:15:51.4978518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 29%] 2024-08-06T21:15:51.4979790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 29%] 2024-08-06T21:15:51.4981046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 29%] 2024-08-06T21:15:51.4982351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 29%] 2024-08-06T21:15:51.4983656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 29%] 2024-08-06T21:15:51.4984916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 29%] 2024-08-06T21:15:51.4986165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 29%] 2024-08-06T21:15:51.4987457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 29%] 2024-08-06T21:15:51.4988804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 29%] 2024-08-06T21:15:51.4990057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 29%] 2024-08-06T21:15:51.4991334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 29%] 2024-08-06T21:15:51.4992602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0089s] [ 29%] 2024-08-06T21:15:51.4993882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0089s] [ 29%] 2024-08-06T21:15:51.4995140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 29%] 2024-08-06T21:15:51.4996428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 30%] 2024-08-06T21:15:51.4997699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0099s] [ 30%] 2024-08-06T21:15:51.4998983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0098s] [ 30%] 2024-08-06T21:15:51.5000305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 30%] 2024-08-06T21:15:51.5001624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 30%] 2024-08-06T21:15:51.5002946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0100s] [ 30%] 2024-08-06T21:15:51.5004580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0099s] [ 30%] 2024-08-06T21:15:51.5005963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0067s] [ 30%] 2024-08-06T21:15:51.5007317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 30%] 2024-08-06T21:15:51.5008606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0110s] [ 30%] 2024-08-06T21:15:51.5009875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0109s] [ 30%] 2024-08-06T21:15:51.5011157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0074s] [ 30%] 2024-08-06T21:15:51.5012430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 30%] 2024-08-06T21:15:51.5013688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0121s] [ 30%] 2024-08-06T21:15:51.5014973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0122s] [ 30%] 2024-08-06T21:15:51.5016239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0088s] [ 30%] 2024-08-06T21:15:51.5017525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 30%] 2024-08-06T21:15:51.5018807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0134s] [ 30%] 2024-08-06T21:15:51.5020164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0134s] [ 30%] 2024-08-06T21:15:51.5021498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0094s] [ 30%] 2024-08-06T21:15:51.5022784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0094s] [ 30%] 2024-08-06T21:15:51.5024040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0083s] [ 30%] 2024-08-06T21:15:51.5025350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0085s] [ 30%] 2024-08-06T21:15:51.5026677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 30%] 2024-08-06T21:15:51.5027939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 30%] 2024-08-06T21:15:51.5029242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0092s] [ 30%] 2024-08-06T21:15:51.5030517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0092s] [ 30%] 2024-08-06T21:15:51.5031792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 30%] 2024-08-06T21:15:51.5033053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 30%] 2024-08-06T21:15:51.5034332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 30%] 2024-08-06T21:15:51.5035601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 30%] 2024-08-06T21:15:51.5036872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 30%] 2024-08-06T21:15:51.5038175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 30%] 2024-08-06T21:15:51.5039498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 30%] 2024-08-06T21:15:51.5040784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 30%] 2024-08-06T21:15:51.5042042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 30%] 2024-08-06T21:15:51.5043364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 30%] 2024-08-06T21:15:51.5044675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 30%] 2024-08-06T21:15:51.5045954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 30%] 2024-08-06T21:15:51.5047206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 30%] 2024-08-06T21:15:51.5048494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 30%] 2024-08-06T21:15:51.5049764Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 30%] 2024-08-06T21:15:51.5051029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 30%] 2024-08-06T21:15:51.5052303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 30%] 2024-08-06T21:15:51.5053574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 30%] 2024-08-06T21:15:51.5054847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 30%] 2024-08-06T21:15:51.5056103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 30%] 2024-08-06T21:15:51.5057416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 30%] 2024-08-06T21:15:51.5058753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 30%] 2024-08-06T21:15:51.5060033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0071s] [ 30%] 2024-08-06T21:15:51.5061297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 30%] 2024-08-06T21:15:51.5062613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 30%] 2024-08-06T21:15:51.5063933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 30%] 2024-08-06T21:15:51.5065183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 30%] 2024-08-06T21:15:51.5066459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 30%] 2024-08-06T21:15:51.5067717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 30%] 2024-08-06T21:15:51.5069038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 30%] 2024-08-06T21:15:51.5070294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 30%] 2024-08-06T21:15:51.5071571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 30%] 2024-08-06T21:15:51.5072832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 30%] 2024-08-06T21:15:51.5074107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 30%] 2024-08-06T21:15:51.5075359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 30%] 2024-08-06T21:15:51.5076664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 30%] 2024-08-06T21:15:51.5077980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 30%] 2024-08-06T21:15:51.5079252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 30%] 2024-08-06T21:15:51.5080531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 30%] 2024-08-06T21:15:51.5081848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 30%] 2024-08-06T21:15:51.5083166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 30%] 2024-08-06T21:15:51.5084425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 30%] 2024-08-06T21:15:51.5085698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 30%] 2024-08-06T21:15:51.5086961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 30%] 2024-08-06T21:15:51.5088214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 30%] 2024-08-06T21:15:51.5089490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 30%] 2024-08-06T21:15:51.5090747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 30%] 2024-08-06T21:15:51.5092031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 30%] 2024-08-06T21:15:51.5093281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 30%] 2024-08-06T21:15:51.5094600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 30%] 2024-08-06T21:15:51.5095905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 30%] 2024-08-06T21:15:51.5097178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 31%] 2024-08-06T21:15:51.5098469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 31%] 2024-08-06T21:15:51.5099790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 31%] 2024-08-06T21:15:51.5101097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 31%] 2024-08-06T21:15:51.5102354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 31%] 2024-08-06T21:15:51.5103666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 31%] 2024-08-06T21:15:51.5105290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 31%] 2024-08-06T21:15:51.5106577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 31%] 2024-08-06T21:15:51.5107825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 31%] 2024-08-06T21:15:51.5109116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 31%] 2024-08-06T21:15:51.5116265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 31%] 2024-08-06T21:15:51.5117663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 31%] 2024-08-06T21:15:51.5118926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 31%] 2024-08-06T21:15:51.5120326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 31%] 2024-08-06T21:15:51.5121669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 31%] 2024-08-06T21:15:51.5122946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0190s] [ 31%] 2024-08-06T21:15:51.5124223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0203s] [ 31%] 2024-08-06T21:15:51.5125569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0192s] [ 31%] 2024-08-06T21:15:51.5126910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0192s] [ 31%] 2024-08-06T21:15:51.5128170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0297s] [ 31%] 2024-08-06T21:15:51.5129446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0304s] [ 31%] 2024-08-06T21:15:51.5130717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0231s] [ 31%] 2024-08-06T21:15:51.5131993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0233s] [ 31%] 2024-08-06T21:15:51.5133254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0204s] [ 31%] 2024-08-06T21:15:51.5134543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0220s] [ 31%] 2024-08-06T21:15:51.5135823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0218s] [ 31%] 2024-08-06T21:15:51.5137095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0218s] [ 31%] 2024-08-06T21:15:51.5138405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0312s] [ 31%] 2024-08-06T21:15:51.5139729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0318s] [ 31%] 2024-08-06T21:15:51.5141008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0259s] [ 31%] 2024-08-06T21:15:51.5142279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.1955s] [ 31%] 2024-08-06T21:15:51.5143595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0242s] [ 31%] 2024-08-06T21:15:51.5144935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0250s] [ 31%] 2024-08-06T21:15:51.5146199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0251s] [ 31%] 2024-08-06T21:15:51.5147467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0250s] [ 31%] 2024-08-06T21:15:51.5148757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0339s] [ 31%] 2024-08-06T21:15:51.5150032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0349s] [ 31%] 2024-08-06T21:15:51.5151291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0308s] [ 31%] 2024-08-06T21:15:51.5152566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0308s] [ 31%] 2024-08-06T21:15:51.5153832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0183s] [ 31%] 2024-08-06T21:15:51.5155121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0193s] [ 31%] 2024-08-06T21:15:51.5156377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0192s] [ 31%] 2024-08-06T21:15:51.5157701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0191s] [ 31%] 2024-08-06T21:15:51.5159017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0298s] [ 31%] 2024-08-06T21:15:51.5160290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0301s] [ 31%] 2024-08-06T21:15:51.5161550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0226s] [ 31%] 2024-08-06T21:15:51.5162872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0227s] [ 31%] 2024-08-06T21:15:51.5164224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0357s] [ 31%] 2024-08-06T21:15:51.5165502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0383s] [ 31%] 2024-08-06T21:15:51.5166768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0352s] [ 31%] 2024-08-06T21:15:51.5168038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.1983s] [ 31%] 2024-08-06T21:15:51.5169309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0576s] [ 31%] 2024-08-06T21:15:51.5170576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0587s] [ 31%] 2024-08-06T21:15:51.5171855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0432s] [ 31%] 2024-08-06T21:15:51.5173121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0433s] [ 31%] 2024-08-06T21:15:51.5174411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0379s] [ 31%] 2024-08-06T21:15:51.5175726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0412s] [ 31%] 2024-08-06T21:15:51.5177043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0417s] [ 31%] 2024-08-06T21:15:51.5178320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0418s] [ 31%] 2024-08-06T21:15:51.5179584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0594s] [ 31%] 2024-08-06T21:15:51.5180914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0602s] [ 31%] 2024-08-06T21:15:51.5182234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0476s] [ 31%] 2024-08-06T21:15:51.5183512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.2096s] [ 31%] 2024-08-06T21:15:51.5184794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0431s] [ 31%] 2024-08-06T21:15:51.5186073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0469s] [ 31%] 2024-08-06T21:15:51.5187338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0478s] [ 31%] 2024-08-06T21:15:51.5188610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0464s] [ 31%] 2024-08-06T21:15:51.5189876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0641s] [ 31%] 2024-08-06T21:15:51.5191149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0651s] [ 31%] 2024-08-06T21:15:51.5192416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0552s] [ 31%] 2024-08-06T21:15:51.5193727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0551s] [ 31%] 2024-08-06T21:15:51.5195054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0336s] [ 31%] 2024-08-06T21:15:51.5196323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0356s] [ 31%] 2024-08-06T21:15:51.5197585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.1948s] [ 31%] 2024-08-06T21:15:51.5198886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0335s] [ 31%] 2024-08-06T21:15:51.5200211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0559s] [ 31%] 2024-08-06T21:15:51.5201477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0571s] [ 31%] 2024-08-06T21:15:51.5202804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0410s] [ 32%] 2024-08-06T21:15:51.5204565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0412s] [ 32%] 2024-08-06T21:15:51.5205867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0120s] [ 32%] 2024-08-06T21:15:51.5207184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0120s] [ 32%] 2024-08-06T21:15:51.5208447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0115s] [ 32%] 2024-08-06T21:15:51.5209722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0115s] [ 32%] 2024-08-06T21:15:51.5210986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0168s] [ 32%] 2024-08-06T21:15:51.5212256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0169s] [ 32%] 2024-08-06T21:15:51.5213605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0130s] [ 32%] 2024-08-06T21:15:51.5214948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0131s] [ 32%] 2024-08-06T21:15:51.5216208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0127s] [ 32%] 2024-08-06T21:15:51.5217465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0129s] [ 32%] 2024-08-06T21:15:51.5218797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0131s] [ 32%] 2024-08-06T21:15:51.5220120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0134s] [ 32%] 2024-08-06T21:15:51.5221385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0179s] [ 32%] 2024-08-06T21:15:51.5222654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0181s] [ 32%] 2024-08-06T21:15:51.5223923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0143s] [ 32%] 2024-08-06T21:15:51.5225206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0144s] [ 32%] 2024-08-06T21:15:51.5226469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0153s] [ 32%] 2024-08-06T21:15:51.5227737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0155s] [ 32%] 2024-08-06T21:15:51.5229016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0163s] [ 32%] 2024-08-06T21:15:51.5230281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0165s] [ 32%] 2024-08-06T21:15:51.5231594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0203s] [ 32%] 2024-08-06T21:15:51.5232945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0205s] [ 32%] 2024-08-06T21:15:51.5234214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.1816s] [ 32%] 2024-08-06T21:15:51.5235484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0176s] [ 32%] 2024-08-06T21:15:51.5236781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0111s] [ 32%] 2024-08-06T21:15:51.5238104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0109s] [ 32%] 2024-08-06T21:15:51.5239361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0112s] [ 32%] 2024-08-06T21:15:51.5240630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0110s] [ 32%] 2024-08-06T21:15:51.5241895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0165s] [ 32%] 2024-08-06T21:15:51.5243173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0164s] [ 32%] 2024-08-06T21:15:51.5244444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0121s] [ 32%] 2024-08-06T21:15:51.5245715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0120s] [ 32%] 2024-08-06T21:15:51.5247013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0105s] [ 32%] 2024-08-06T21:15:51.5248277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0104s] [ 32%] 2024-08-06T21:15:51.5249540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0094s] [ 32%] 2024-08-06T21:15:51.5250841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0095s] [ 32%] 2024-08-06T21:15:51.5252176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0146s] [ 32%] 2024-08-06T21:15:51.5253436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0149s] [ 32%] 2024-08-06T21:15:51.5254725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0109s] [ 32%] 2024-08-06T21:15:51.5256030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0106s] [ 32%] 2024-08-06T21:15:51.5257336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0114s] [ 32%] 2024-08-06T21:15:51.5258602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0110s] [ 32%] 2024-08-06T21:15:51.5259860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0108s] [ 32%] 2024-08-06T21:15:51.5261147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0107s] [ 32%] 2024-08-06T21:15:51.5262401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0153s] [ 32%] 2024-08-06T21:15:51.5263669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0154s] [ 32%] 2024-08-06T21:15:51.5264941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0116s] [ 32%] 2024-08-06T21:15:51.5266217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0113s] [ 32%] 2024-08-06T21:15:51.5267468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0139s] [ 32%] 2024-08-06T21:15:51.5268772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0138s] [ 32%] 2024-08-06T21:15:51.5270082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0138s] [ 32%] 2024-08-06T21:15:51.5271609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0136s] [ 32%] 2024-08-06T21:15:51.5273243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0172s] [ 32%] 2024-08-06T21:15:51.5275763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0174s] [ 32%] 2024-08-06T21:15:51.5278207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0139s] [ 32%] 2024-08-06T21:15:51.5280592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0139s] [ 32%] 2024-08-06T21:15:51.5282977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0094s] [ 32%] 2024-08-06T21:15:51.5285379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0094s] [ 32%] 2024-08-06T21:15:51.5287771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0085s] [ 32%] 2024-08-06T21:15:51.5290149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 32%] 2024-08-06T21:15:51.5292512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0134s] [ 32%] 2024-08-06T21:15:51.5294901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0134s] [ 32%] 2024-08-06T21:15:51.5297284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0091s] [ 32%] 2024-08-06T21:15:51.5299661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0093s] [ 32%] 2024-08-06T21:15:51.5302085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0683s] [ 32%] 2024-08-06T21:15:51.5304986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0743s] [ 32%] 2024-08-06T21:15:51.5307411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0648s] [ 32%] 2024-08-06T21:15:51.5309799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.2153s] [ 32%] 2024-08-06T21:15:51.5312303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.1105s] [ 32%] 2024-08-06T21:15:51.5314782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.1147s] [ 32%] 2024-08-06T21:15:51.5317171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0806s] [ 32%] 2024-08-06T21:15:51.5319564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0812s] [ 32%] 2024-08-06T21:15:51.5321973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0728s] [ 32%] 2024-08-06T21:15:51.5324367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0796s] [ 32%] 2024-08-06T21:15:51.5326759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.2311s] [ 32%] 2024-08-06T21:15:51.5329153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0732s] [ 32%] 2024-08-06T21:15:51.5331550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.1131s] [ 33%] 2024-08-06T21:15:51.5333942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.1173s] [ 33%] 2024-08-06T21:15:51.5336398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0897s] [ 33%] 2024-08-06T21:15:51.5338874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.2327s] [ 33%] 2024-08-06T21:15:51.5341266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0816s] [ 33%] 2024-08-06T21:15:51.5343645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0885s] [ 33%] 2024-08-06T21:15:51.5346068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0864s] [ 33%] 2024-08-06T21:15:51.5348516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0855s] [ 33%] 2024-08-06T21:15:51.5350908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.1219s] [ 33%] 2024-08-06T21:15:51.5353332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.2827s] [ 33%] 2024-08-06T21:15:51.5355752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.1029s] [ 33%] 2024-08-06T21:15:51.5358153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.1036s] [ 33%] 2024-08-06T21:15:51.5360534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0663s] [ 33%] 2024-08-06T21:15:51.5362949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0707s] [ 33%] 2024-08-06T21:15:51.5365342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0612s] [ 33%] 2024-08-06T21:15:51.5367738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.2117s] [ 33%] 2024-08-06T21:15:51.5370153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.1086s] [ 33%] 2024-08-06T21:15:51.5372602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.1125s] [ 33%] 2024-08-06T21:15:51.5375053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0769s] [ 33%] 2024-08-06T21:15:51.5377465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0770s] [ 33%] 2024-08-06T21:15:51.5379860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0127s] [ 33%] 2024-08-06T21:15:51.5382288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0129s] [ 33%] 2024-08-06T21:15:51.5384736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0130s] [ 33%] 2024-08-06T21:15:51.5387109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0129s] [ 33%] 2024-08-06T21:15:51.5389504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0182s] [ 33%] 2024-08-06T21:15:51.5391901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0182s] [ 33%] 2024-08-06T21:15:51.5394288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0148s] [ 33%] 2024-08-06T21:15:51.5396676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0149s] [ 33%] 2024-08-06T21:15:51.5399079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0134s] [ 33%] 2024-08-06T21:15:51.5401465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0137s] [ 33%] 2024-08-06T21:15:51.5403882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.1787s] [ 33%] 2024-08-06T21:15:51.5406806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0147s] [ 33%] 2024-08-06T21:15:51.5409265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0188s] [ 33%] 2024-08-06T21:15:51.5411642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0188s] [ 33%] 2024-08-06T21:15:51.5414016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0164s] [ 33%] 2024-08-06T21:15:51.5416456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0165s] [ 33%] 2024-08-06T21:15:51.5418903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0154s] [ 33%] 2024-08-06T21:15:51.5421276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0162s] [ 33%] 2024-08-06T21:15:51.5423654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0178s] [ 33%] 2024-08-06T21:15:51.5426036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0172s] [ 33%] 2024-08-06T21:15:51.5428419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0211s] [ 33%] 2024-08-06T21:15:51.5430795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0210s] [ 33%] 2024-08-06T21:15:51.5433186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0195s] [ 33%] 2024-08-06T21:15:51.5435578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0195s] [ 33%] 2024-08-06T21:15:51.5437971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0120s] [ 33%] 2024-08-06T21:15:51.5440359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0123s] [ 33%] 2024-08-06T21:15:51.5442784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0121s] [ 33%] 2024-08-06T21:15:51.5445217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0121s] [ 33%] 2024-08-06T21:15:51.5447617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0173s] [ 33%] 2024-08-06T21:15:51.5450044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0174s] [ 33%] 2024-08-06T21:15:51.5452475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0139s] [ 33%] 2024-08-06T21:15:51.5454900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0144s] [ 33%] 2024-08-06T21:15:51.5457281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0107s] [ 33%] 2024-08-06T21:15:51.5459659Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0106s] [ 33%] 2024-08-06T21:15:51.5462062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0097s] [ 33%] 2024-08-06T21:15:51.5464462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0094s] [ 33%] 2024-08-06T21:15:51.5466864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0148s] [ 33%] 2024-08-06T21:15:51.5469291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0158s] [ 33%] 2024-08-06T21:15:51.5471674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0106s] [ 33%] 2024-08-06T21:15:51.5474074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0106s] [ 33%] 2024-08-06T21:15:51.5476523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0113s] [ 33%] 2024-08-06T21:15:51.5478973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0113s] [ 33%] 2024-08-06T21:15:51.5481392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0100s] [ 33%] 2024-08-06T21:15:51.5483789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0104s] [ 33%] 2024-08-06T21:15:51.5486214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0145s] [ 33%] 2024-08-06T21:15:51.5488682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0148s] [ 33%] 2024-08-06T21:15:51.5491078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0105s] [ 33%] 2024-08-06T21:15:51.5493477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0105s] [ 33%] 2024-08-06T21:15:51.5495899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0132s] [ 33%] 2024-08-06T21:15:51.5498283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0133s] [ 33%] 2024-08-06T21:15:51.5500663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0130s] [ 33%] 2024-08-06T21:15:51.5503099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0132s] [ 33%] 2024-08-06T21:15:51.5505902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0170s] [ 33%] 2024-08-06T21:15:51.5508343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0170s] [ 33%] 2024-08-06T21:15:51.5510734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0135s] [ 33%] 2024-08-06T21:15:51.5513205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0134s] [ 33%] 2024-08-06T21:15:51.5515680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0092s] [ 33%] 2024-08-06T21:15:51.5518086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0089s] [ 34%] 2024-08-06T21:15:51.5520531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0082s] [ 34%] 2024-08-06T21:15:51.5522985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 34%] 2024-08-06T21:15:51.5525387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0132s] [ 34%] 2024-08-06T21:15:51.5527780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0132s] [ 34%] 2024-08-06T21:15:51.5530180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0086s] [ 34%] 2024-08-06T21:15:51.5532604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0089s] [ 34%] 2024-08-06T21:15:51.5535006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 34%] 2024-08-06T21:15:51.5537430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 34%] 2024-08-06T21:15:51.5539839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 34%] 2024-08-06T21:15:51.5542243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 34%] 2024-08-06T21:15:51.5544644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 34%] 2024-08-06T21:15:51.5547096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 34%] 2024-08-06T21:15:51.5550118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 34%] 2024-08-06T21:15:51.5552502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 34%] 2024-08-06T21:15:51.5554902Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 34%] 2024-08-06T21:15:51.5557362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 34%] 2024-08-06T21:15:51.5559811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 34%] 2024-08-06T21:15:51.5562212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 34%] 2024-08-06T21:15:51.5564601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0082s] [ 34%] 2024-08-06T21:15:51.5567013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 34%] 2024-08-06T21:15:51.5569415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0080s] [ 34%] 2024-08-06T21:15:51.5571816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 34%] 2024-08-06T21:15:51.5574250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0077s] [ 34%] 2024-08-06T21:15:51.5576657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0078s] [ 34%] 2024-08-06T21:15:51.5579038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0078s] [ 34%] 2024-08-06T21:15:51.5581432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 34%] 2024-08-06T21:15:51.5583890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0093s] [ 34%] 2024-08-06T21:15:51.5586359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0095s] [ 34%] 2024-08-06T21:15:51.5588770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0087s] [ 34%] 2024-08-06T21:15:51.5591208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0087s] [ 34%] 2024-08-06T21:15:51.5593668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 34%] 2024-08-06T21:15:51.5596072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 34%] 2024-08-06T21:15:51.5598473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 34%] 2024-08-06T21:15:51.5600901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 34%] 2024-08-06T21:15:51.5603371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 34%] 2024-08-06T21:15:51.5606180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0078s] [ 34%] 2024-08-06T21:15:51.5608587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 34%] 2024-08-06T21:15:51.5611014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 34%] 2024-08-06T21:15:51.5613416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0084s] [ 34%] 2024-08-06T21:15:51.5615807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0087s] [ 34%] 2024-08-06T21:15:51.5618297Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0081s] [ 34%] 2024-08-06T21:15:51.5620768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0083s] [ 34%] 2024-08-06T21:15:51.5623164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0107s] [ 34%] 2024-08-06T21:15:51.5625596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0109s] [ 34%] 2024-08-06T21:15:51.5628143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0092s] [ 34%] 2024-08-06T21:15:51.5630625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0092s] [ 34%] 2024-08-06T21:15:51.5633006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0086s] [ 34%] 2024-08-06T21:15:51.5635422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0092s] [ 34%] 2024-08-06T21:15:51.5637817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0091s] [ 34%] 2024-08-06T21:15:51.5640206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0092s] [ 34%] 2024-08-06T21:15:51.5642600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0114s] [ 34%] 2024-08-06T21:15:51.5644995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0114s] [ 34%] 2024-08-06T21:15:51.5647409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0102s] [ 34%] 2024-08-06T21:15:51.5649804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0102s] [ 34%] 2024-08-06T21:15:51.5652271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0100s] [ 34%] 2024-08-06T21:15:51.5654724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0108s] [ 34%] 2024-08-06T21:15:51.5657122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0107s] [ 34%] 2024-08-06T21:15:51.5659501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0108s] [ 34%] 2024-08-06T21:15:51.5661948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0126s] [ 34%] 2024-08-06T21:15:51.5664408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0128s] [ 34%] 2024-08-06T21:15:51.5666815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0115s] [ 34%] 2024-08-06T21:15:51.5669219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0117s] [ 34%] 2024-08-06T21:15:51.5671623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0082s] [ 34%] 2024-08-06T21:15:51.5674042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 34%] 2024-08-06T21:15:51.5676433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0079s] [ 34%] 2024-08-06T21:15:51.5678830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 34%] 2024-08-06T21:15:51.5681251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0104s] [ 34%] 2024-08-06T21:15:51.5683647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0107s] [ 34%] 2024-08-06T21:15:51.5686018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0088s] [ 34%] 2024-08-06T21:15:51.5688481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0088s] [ 34%] 2024-08-06T21:15:51.5690941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 34%] 2024-08-06T21:15:51.5693342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 34%] 2024-08-06T21:15:51.5695749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 34%] 2024-08-06T21:15:51.5698174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 34%] 2024-08-06T21:15:51.5700622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0068s] [ 34%] 2024-08-06T21:15:51.5703148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 34%] 2024-08-06T21:15:51.5705946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 35%] 2024-08-06T21:15:51.5708359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 35%] 2024-08-06T21:15:51.5710768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 35%] 2024-08-06T21:15:51.5713140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 35%] 2024-08-06T21:15:51.5715533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 35%] 2024-08-06T21:15:51.5717939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 35%] 2024-08-06T21:15:51.5720336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 35%] 2024-08-06T21:15:51.5722813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 35%] 2024-08-06T21:15:51.5725270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 35%] 2024-08-06T21:15:51.5727675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 35%] 2024-08-06T21:15:51.5730076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 35%] 2024-08-06T21:15:51.5732542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 35%] 2024-08-06T21:15:51.5735010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 35%] 2024-08-06T21:15:51.5737406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 35%] 2024-08-06T21:15:51.5739811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 35%] 2024-08-06T21:15:51.5742219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0070s] [ 35%] 2024-08-06T21:15:51.5744625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 35%] 2024-08-06T21:15:51.5747023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 35%] 2024-08-06T21:15:51.5749413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 35%] 2024-08-06T21:15:51.5751797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 35%] 2024-08-06T21:15:51.5754179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 35%] 2024-08-06T21:15:51.5756588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 35%] 2024-08-06T21:15:51.5759052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 35%] 2024-08-06T21:15:51.5761513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 35%] 2024-08-06T21:15:51.5763900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 35%] 2024-08-06T21:15:51.5766292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 35%] 2024-08-06T21:15:51.5768730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 35%] 2024-08-06T21:15:51.5771172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 35%] 2024-08-06T21:15:51.5773565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 35%] 2024-08-06T21:15:51.5775979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 35%] 2024-08-06T21:15:51.5778362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 35%] 2024-08-06T21:15:51.5780747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 35%] 2024-08-06T21:15:51.5783118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 35%] 2024-08-06T21:15:51.5785506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 35%] 2024-08-06T21:15:51.5787901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 35%] 2024-08-06T21:15:51.5790262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 35%] 2024-08-06T21:15:51.5792633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0068s] [ 35%] 2024-08-06T21:15:51.5795050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 35%] 2024-08-06T21:15:51.5797487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0068s] [ 35%] 2024-08-06T21:15:51.5799904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 35%] 2024-08-06T21:15:51.5802343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 35%] 2024-08-06T21:15:51.5805251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 35%] 2024-08-06T21:15:51.5807705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 35%] 2024-08-06T21:15:51.5810089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 35%] 2024-08-06T21:15:51.5812475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 35%] 2024-08-06T21:15:51.5814867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 35%] 2024-08-06T21:15:51.5817234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0068s] [ 35%] 2024-08-06T21:15:51.5819646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 35%] 2024-08-06T21:15:51.5822067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 35%] 2024-08-06T21:15:51.5824465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 35%] 2024-08-06T21:15:51.5826841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 35%] 2024-08-06T21:15:51.5829313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 35%] 2024-08-06T21:15:51.5831741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 35%] 2024-08-06T21:15:51.5834126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 35%] 2024-08-06T21:15:51.5836516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 35%] 2024-08-06T21:15:51.5838951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 35%] 2024-08-06T21:15:51.5841403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 35%] 2024-08-06T21:15:51.5843761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 35%] 2024-08-06T21:15:51.5846148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0130s] [ 35%] 2024-08-06T21:15:51.5848551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0138s] [ 35%] 2024-08-06T21:15:51.5850957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0124s] [ 35%] 2024-08-06T21:15:51.5853345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0122s] [ 35%] 2024-08-06T21:15:51.5855750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0178s] [ 35%] 2024-08-06T21:15:51.5858145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0184s] [ 35%] 2024-08-06T21:15:51.5860551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0142s] [ 35%] 2024-08-06T21:15:51.5862959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0138s] [ 35%] 2024-08-06T21:15:51.5865404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0139s] [ 35%] 2024-08-06T21:15:51.5867881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0147s] [ 35%] 2024-08-06T21:15:51.5870259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0139s] [ 35%] 2024-08-06T21:15:51.5872695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0138s] [ 35%] 2024-08-06T21:15:51.5875140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0188s] [ 35%] 2024-08-06T21:15:51.5877546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0192s] [ 35%] 2024-08-06T21:15:51.5879951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0157s] [ 35%] 2024-08-06T21:15:51.5882360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0159s] [ 35%] 2024-08-06T21:15:51.5884738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0163s] [ 35%] 2024-08-06T21:15:51.5887140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0170s] [ 35%] 2024-08-06T21:15:51.5889529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0164s] [ 35%] 2024-08-06T21:15:51.5891946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0164s] [ 35%] 2024-08-06T21:15:51.5894350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0211s] [ 36%] 2024-08-06T21:15:51.5896738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0214s] [ 36%] 2024-08-06T21:15:51.5899227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0188s] [ 36%] 2024-08-06T21:15:51.5901700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0185s] [ 36%] 2024-08-06T21:15:51.5904525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0122s] [ 36%] 2024-08-06T21:15:51.5906988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0129s] [ 36%] 2024-08-06T21:15:51.5909475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0113s] [ 36%] 2024-08-06T21:15:51.5911937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0115s] [ 36%] 2024-08-06T21:15:51.5914373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0174s] [ 36%] 2024-08-06T21:15:51.5916790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0178s] [ 36%] 2024-08-06T21:15:51.5919232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0132s] [ 36%] 2024-08-06T21:15:51.5921655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0133s] [ 36%] 2024-08-06T21:15:51.5924036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 36%] 2024-08-06T21:15:51.5926461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 36%] 2024-08-06T21:15:51.5928870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 36%] 2024-08-06T21:15:51.5931267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 36%] 2024-08-06T21:15:51.5933664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 36%] 2024-08-06T21:15:51.5936108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 36%] 2024-08-06T21:15:51.5938585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 36%] 2024-08-06T21:15:51.5940981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 36%] 2024-08-06T21:15:51.5943434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 36%] 2024-08-06T21:15:51.5945889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 36%] 2024-08-06T21:15:51.5948286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 36%] 2024-08-06T21:15:51.5950670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 36%] 2024-08-06T21:15:51.5953075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 36%] 2024-08-06T21:15:51.5955478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 36%] 2024-08-06T21:15:51.5957875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 36%] 2024-08-06T21:15:51.5960276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 36%] 2024-08-06T21:15:51.5962686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 36%] 2024-08-06T21:15:51.5965090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 36%] 2024-08-06T21:15:51.5967481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0064s] [ 36%] 2024-08-06T21:15:51.5969938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 36%] 2024-08-06T21:15:51.5972389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0068s] [ 36%] 2024-08-06T21:15:51.5974787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 36%] 2024-08-06T21:15:51.5977186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0073s] [ 36%] 2024-08-06T21:15:51.5979633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 36%] 2024-08-06T21:15:51.5982098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 36%] 2024-08-06T21:15:51.5984506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 36%] 2024-08-06T21:15:51.5986892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 36%] 2024-08-06T21:15:51.5989264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 36%] 2024-08-06T21:15:51.5991670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 36%] 2024-08-06T21:15:51.5994055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 36%] 2024-08-06T21:15:51.5996451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 36%] 2024-08-06T21:15:51.5998849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 36%] 2024-08-06T21:15:51.6001251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 36%] 2024-08-06T21:15:51.6003658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 36%] 2024-08-06T21:15:51.6006546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 36%] 2024-08-06T21:15:51.6009009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 36%] 2024-08-06T21:15:51.6011405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 36%] 2024-08-06T21:15:51.6013854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 36%] 2024-08-06T21:15:51.6016316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 36%] 2024-08-06T21:15:51.6018735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 36%] 2024-08-06T21:15:51.6021125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 36%] 2024-08-06T21:15:51.6023518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 36%] 2024-08-06T21:15:51.6025913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 36%] 2024-08-06T21:15:51.6028300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 36%] 2024-08-06T21:15:51.6030673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 36%] 2024-08-06T21:15:51.6033082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 36%] 2024-08-06T21:15:51.6035481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 36%] 2024-08-06T21:15:51.6037899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 36%] 2024-08-06T21:15:51.6040356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 36%] 2024-08-06T21:15:51.6042779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 36%] 2024-08-06T21:15:51.6045165Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 36%] 2024-08-06T21:15:51.6047572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 36%] 2024-08-06T21:15:51.6050004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0069s] [ 36%] 2024-08-06T21:15:51.6052449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 36%] 2024-08-06T21:15:51.6054837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 36%] 2024-08-06T21:15:51.6057252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 36%] 2024-08-06T21:15:51.6059644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 36%] 2024-08-06T21:15:51.6062030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 36%] 2024-08-06T21:15:51.6064408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 36%] 2024-08-06T21:15:51.6066791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 36%] 2024-08-06T21:15:51.6069166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0069s] [ 36%] 2024-08-06T21:15:51.6071553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 36%] 2024-08-06T21:15:51.6073933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 36%] 2024-08-06T21:15:51.6076377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 36%] 2024-08-06T21:15:51.6078827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 36%] 2024-08-06T21:15:51.6081203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 37%] 2024-08-06T21:15:51.6083600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 37%] 2024-08-06T21:15:51.6086041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 37%] 2024-08-06T21:15:51.6088484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 37%] 2024-08-06T21:15:51.6090881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 37%] 2024-08-06T21:15:51.6093306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 37%] 2024-08-06T21:15:51.6095701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 37%] 2024-08-06T21:15:51.6098149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 37%] 2024-08-06T21:15:51.6100550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 37%] 2024-08-06T21:15:51.6103001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 37%] 2024-08-06T21:15:51.6105783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 37%] 2024-08-06T21:15:51.6108232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 37%] 2024-08-06T21:15:51.6110748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 37%] 2024-08-06T21:15:51.6113222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 37%] 2024-08-06T21:15:51.6115626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 37%] 2024-08-06T21:15:51.6118054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0067s] [ 37%] 2024-08-06T21:15:51.6120520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 37%] 2024-08-06T21:15:51.6122972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0064s] [ 37%] 2024-08-06T21:15:51.6125385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 37%] 2024-08-06T21:15:51.6127807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0068s] [ 37%] 2024-08-06T21:15:51.6130219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 37%] 2024-08-06T21:15:51.6132623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 37%] 2024-08-06T21:15:51.6135011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 37%] 2024-08-06T21:15:51.6137412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 37%] 2024-08-06T21:15:51.6139819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 37%] 2024-08-06T21:15:51.6142209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 37%] 2024-08-06T21:15:51.6144603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 37%] 2024-08-06T21:15:51.6147076Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 37%] 2024-08-06T21:15:51.6149515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 37%] 2024-08-06T21:15:51.6151909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 37%] 2024-08-06T21:15:51.6154307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 37%] 2024-08-06T21:15:51.6156776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0068s] [ 37%] 2024-08-06T21:15:51.6159226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0070s] [ 37%] 2024-08-06T21:15:51.6161601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 37%] 2024-08-06T21:15:51.6164014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 37%] 2024-08-06T21:15:51.6166429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0077s] [ 37%] 2024-08-06T21:15:51.6168835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 37%] 2024-08-06T21:15:51.6171241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 37%] 2024-08-06T21:15:51.6173649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 37%] 2024-08-06T21:15:51.6176064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0073s] [ 37%] 2024-08-06T21:15:51.6178465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 37%] 2024-08-06T21:15:51.6180931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0064s] [ 37%] 2024-08-06T21:15:51.6183461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 37%] 2024-08-06T21:15:51.6185856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0082s] [ 37%] 2024-08-06T21:15:51.6188261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0081s] [ 37%] 2024-08-06T21:15:51.6190704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 37%] 2024-08-06T21:15:51.6193178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 37%] 2024-08-06T21:15:51.6195592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0086s] [ 37%] 2024-08-06T21:15:51.6197999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0087s] [ 37%] 2024-08-06T21:15:51.6200387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0075s] [ 37%] 2024-08-06T21:15:51.6202791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 37%] 2024-08-06T21:15:51.6205622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0095s] [ 37%] 2024-08-06T21:15:51.6208043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0095s] [ 37%] 2024-08-06T21:15:51.6210452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0082s] [ 37%] 2024-08-06T21:15:51.6212877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0083s] [ 37%] 2024-08-06T21:15:51.6215254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 37%] 2024-08-06T21:15:51.6217728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 37%] 2024-08-06T21:15:51.6220193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 37%] 2024-08-06T21:15:51.6222579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 37%] 2024-08-06T21:15:51.6224975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0073s] [ 37%] 2024-08-06T21:15:51.6227422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 37%] 2024-08-06T21:15:51.6229887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 37%] 2024-08-06T21:15:51.6232301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 37%] 2024-08-06T21:15:51.6234697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 37%] 2024-08-06T21:15:51.6237101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 37%] 2024-08-06T21:15:51.6239483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 37%] 2024-08-06T21:15:51.6241860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 37%] 2024-08-06T21:15:51.6244273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 37%] 2024-08-06T21:15:51.6246680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 37%] 2024-08-06T21:15:51.6249090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 37%] 2024-08-06T21:15:51.6251553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 37%] 2024-08-06T21:15:51.6253976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 37%] 2024-08-06T21:15:51.6256363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 37%] 2024-08-06T21:15:51.6258768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 37%] 2024-08-06T21:15:51.6261198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 37%] 2024-08-06T21:15:51.6263644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 37%] 2024-08-06T21:15:51.6266057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 37%] 2024-08-06T21:15:51.6268446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 37%] 2024-08-06T21:15:51.6270847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 38%] 2024-08-06T21:15:51.6273240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 38%] 2024-08-06T21:15:51.6275626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 38%] 2024-08-06T21:15:51.6278017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 38%] 2024-08-06T21:15:51.6280388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 38%] 2024-08-06T21:15:51.6282778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0068s] [ 38%] 2024-08-06T21:15:51.6285175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 38%] 2024-08-06T21:15:51.6287629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 38%] 2024-08-06T21:15:51.6290092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 38%] 2024-08-06T21:15:51.6292463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 38%] 2024-08-06T21:15:51.6294841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 38%] 2024-08-06T21:15:51.6297258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 38%] 2024-08-06T21:15:51.6299682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 38%] 2024-08-06T21:15:51.6302067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 38%] 2024-08-06T21:15:51.6304861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 38%] 2024-08-06T21:15:51.6307292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 38%] 2024-08-06T21:15:51.6309675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_37_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 38%] 2024-08-06T21:15:51.6312052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 38%] 2024-08-06T21:15:51.6314460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 38%] 2024-08-06T21:15:51.6316847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 38%] 2024-08-06T21:15:51.6319210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 38%] 2024-08-06T21:15:51.6321699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 38%] 2024-08-06T21:15:51.6324160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 38%] 2024-08-06T21:15:51.6326561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 38%] 2024-08-06T21:15:51.6328943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 38%] 2024-08-06T21:15:51.6331388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 38%] 2024-08-06T21:15:51.6333824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 38%] 2024-08-06T21:15:51.6336210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 38%] 2024-08-06T21:15:51.6338590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 38%] 2024-08-06T21:15:51.6340997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 38%] 2024-08-06T21:15:51.6343416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 38%] 2024-08-06T21:15:51.6345785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 38%] 2024-08-06T21:15:51.6348178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 38%] 2024-08-06T21:15:51.6350575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 38%] 2024-08-06T21:15:51.6352966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 38%] 2024-08-06T21:15:51.6355348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 38%] 2024-08-06T21:15:51.6357781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 38%] 2024-08-06T21:15:51.6360207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 38%] 2024-08-06T21:15:51.6362613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 38%] 2024-08-06T21:15:51.6364993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 38%] 2024-08-06T21:15:51.6367432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 38%] 2024-08-06T21:15:51.6369889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 38%] 2024-08-06T21:15:51.6372256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 38%] 2024-08-06T21:15:51.6374637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 38%] 2024-08-06T21:15:51.6377049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 38%] 2024-08-06T21:15:51.6379441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 38%] 2024-08-06T21:15:51.6381829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 38%] 2024-08-06T21:15:51.6384212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 38%] 2024-08-06T21:15:51.6386604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 38%] 2024-08-06T21:15:51.6389001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0097s] [ 38%] 2024-08-06T21:15:51.6391399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0099s] [ 38%] 2024-08-06T21:15:51.6393846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0076s] [ 38%] 2024-08-06T21:15:51.6396303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0076s] [ 38%] 2024-08-06T21:15:51.6398697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0107s] [ 38%] 2024-08-06T21:15:51.6401107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0111s] [ 38%] 2024-08-06T21:15:51.6403627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0083s] [ 38%] 2024-08-06T21:15:51.6406497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 38%] 2024-08-06T21:15:51.6408941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0109s] [ 38%] 2024-08-06T21:15:51.6411332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0112s] [ 38%] 2024-08-06T21:15:51.6413736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0086s] [ 38%] 2024-08-06T21:15:51.6416131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 38%] 2024-08-06T21:15:51.6418555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0121s] [ 38%] 2024-08-06T21:15:51.6420964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0123s] [ 38%] 2024-08-06T21:15:51.6423372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0094s] [ 38%] 2024-08-06T21:15:51.6425756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0093s] [ 38%] 2024-08-06T21:15:51.6428250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0135s] [ 38%] 2024-08-06T21:15:51.6430726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0136s] [ 38%] 2024-08-06T21:15:51.6433135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0107s] [ 38%] 2024-08-06T21:15:51.6435531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0106s] [ 38%] 2024-08-06T21:15:51.6437986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0148s] [ 38%] 2024-08-06T21:15:51.6440478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0150s] [ 38%] 2024-08-06T21:15:51.6442876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0113s] [ 38%] 2024-08-06T21:15:51.6445277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0115s] [ 38%] 2024-08-06T21:15:51.6447706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0092s] [ 38%] 2024-08-06T21:15:51.6450105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0094s] [ 38%] 2024-08-06T21:15:51.6452478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0071s] [ 38%] 2024-08-06T21:15:51.6454897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 38%] 2024-08-06T21:15:51.6457299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0104s] [ 39%] 2024-08-06T21:15:51.6459699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0102s] [ 39%] 2024-08-06T21:15:51.6462094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0079s] [ 39%] 2024-08-06T21:15:51.6464529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0077s] [ 39%] 2024-08-06T21:15:51.6466990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 39%] 2024-08-06T21:15:51.6469383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 39%] 2024-08-06T21:15:51.6471772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 39%] 2024-08-06T21:15:51.6474236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 39%] 2024-08-06T21:15:51.6476677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 39%] 2024-08-06T21:15:51.6479055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 39%] 2024-08-06T21:15:51.6481454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 39%] 2024-08-06T21:15:51.6483855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 39%] 2024-08-06T21:15:51.6486245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 39%] 2024-08-06T21:15:51.6491527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 39%] 2024-08-06T21:15:51.6494786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 39%] 2024-08-06T21:15:51.6496086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 39%] 2024-08-06T21:15:51.6497343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 39%] 2024-08-06T21:15:51.6498665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 39%] 2024-08-06T21:15:51.6499960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 39%] 2024-08-06T21:15:51.6501254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 39%] 2024-08-06T21:15:51.6502509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 39%] 2024-08-06T21:15:51.6503886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 39%] 2024-08-06T21:15:51.6505717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 39%] 2024-08-06T21:15:51.6506984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 39%] 2024-08-06T21:15:51.6508265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 39%] 2024-08-06T21:15:51.6509536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 39%] 2024-08-06T21:15:51.6510825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 39%] 2024-08-06T21:15:51.6512088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 39%] 2024-08-06T21:15:51.6513462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 39%] 2024-08-06T21:15:51.6514802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 39%] 2024-08-06T21:15:51.6516065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 39%] 2024-08-06T21:15:51.6517321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 39%] 2024-08-06T21:15:51.6518580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 39%] 2024-08-06T21:15:51.6519864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 39%] 2024-08-06T21:15:51.6521138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 39%] 2024-08-06T21:15:51.6522418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 39%] 2024-08-06T21:15:51.6523733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 39%] 2024-08-06T21:15:51.6525080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 39%] 2024-08-06T21:15:51.6526328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 39%] 2024-08-06T21:15:51.6527607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 39%] 2024-08-06T21:15:51.6528866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 39%] 2024-08-06T21:15:51.6530197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 39%] 2024-08-06T21:15:51.6531496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 39%] 2024-08-06T21:15:51.6532824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 39%] 2024-08-06T21:15:51.6534163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 39%] 2024-08-06T21:15:51.6535424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 39%] 2024-08-06T21:15:51.6536694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 39%] 2024-08-06T21:15:51.6537954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 39%] 2024-08-06T21:15:51.6539226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 39%] 2024-08-06T21:15:51.6540484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 39%] 2024-08-06T21:15:51.6541794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 39%] 2024-08-06T21:15:51.6543131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 39%] 2024-08-06T21:15:51.6544383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 39%] 2024-08-06T21:15:51.6545654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 39%] 2024-08-06T21:15:51.6546908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 39%] 2024-08-06T21:15:51.6548185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 39%] 2024-08-06T21:15:51.6549437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 39%] 2024-08-06T21:15:51.6550759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 39%] 2024-08-06T21:15:51.6552072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 39%] 2024-08-06T21:15:51.6553351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 39%] 2024-08-06T21:15:51.6554598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 39%] 2024-08-06T21:15:51.6555857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 39%] 2024-08-06T21:15:51.6557126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 39%] 2024-08-06T21:15:51.6558378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 39%] 2024-08-06T21:15:51.6559647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 39%] 2024-08-06T21:15:51.6560959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 39%] 2024-08-06T21:15:51.6562285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 39%] 2024-08-06T21:15:51.6563541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 39%] 2024-08-06T21:15:51.6564327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_mask_variants_mask_dim_1_cuda PASSED [0.0034s] [ 39%] 2024-08-06T21:15:51.6565093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_mask_variants_mask_dim_2_cuda PASSED [0.0019s] [ 39%] 2024-08-06T21:15:51.6565858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_mask_variants_mask_dim_3_cuda PASSED [0.0018s] [ 39%] 2024-08-06T21:15:51.6566635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_mask_variants_mask_dim_4_cuda PASSED [0.0018s] [ 39%] 2024-08-06T21:15:51.6567877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0074s] [ 39%] 2024-08-06T21:15:51.6569178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 39%] 2024-08-06T21:15:51.6570465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 39%] 2024-08-06T21:15:51.6571716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 39%] 2024-08-06T21:15:51.6572961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0096s] [ 39%] 2024-08-06T21:15:51.6574196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0099s] [ 40%] 2024-08-06T21:15:51.6575445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 40%] 2024-08-06T21:15:51.6576678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 40%] 2024-08-06T21:15:51.6577904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0265s] [ 40%] 2024-08-06T21:15:51.6579173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 40%] 2024-08-06T21:15:51.6580459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 40%] 2024-08-06T21:15:51.6581692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 40%] 2024-08-06T21:15:51.6582916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0089s] [ 40%] 2024-08-06T21:15:51.6584162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0091s] [ 40%] 2024-08-06T21:15:51.6585390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 40%] 2024-08-06T21:15:51.6586628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 40%] 2024-08-06T21:15:51.6587894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0077s] [ 40%] 2024-08-06T21:15:51.6589195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 40%] 2024-08-06T21:15:51.6590418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 40%] 2024-08-06T21:15:51.6591680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 40%] 2024-08-06T21:15:51.6592910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0098s] [ 40%] 2024-08-06T21:15:51.6594167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0099s] [ 40%] 2024-08-06T21:15:51.6595393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 40%] 2024-08-06T21:15:51.6596621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 40%] 2024-08-06T21:15:51.6597915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0076s] [ 40%] 2024-08-06T21:15:51.6599197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 40%] 2024-08-06T21:15:51.6600427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 40%] 2024-08-06T21:15:51.6601669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 40%] 2024-08-06T21:15:51.6602914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0093s] [ 40%] 2024-08-06T21:15:51.6604704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0093s] [ 40%] 2024-08-06T21:15:51.6606592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 40%] 2024-08-06T21:15:51.6607928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 40%] 2024-08-06T21:15:51.6609233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0083s] [ 40%] 2024-08-06T21:15:51.6610480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0089s] [ 40%] 2024-08-06T21:15:51.6611702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 40%] 2024-08-06T21:15:51.6612972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 40%] 2024-08-06T21:15:51.6614207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0102s] [ 40%] 2024-08-06T21:15:51.6615457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 40%] 2024-08-06T21:15:51.6616680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 40%] 2024-08-06T21:15:51.6617994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 40%] 2024-08-06T21:15:51.6619283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0079s] [ 40%] 2024-08-06T21:15:51.6620499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 40%] 2024-08-06T21:15:51.6621732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 40%] 2024-08-06T21:15:51.6622975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 40%] 2024-08-06T21:15:51.6624220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0097s] [ 40%] 2024-08-06T21:15:51.6625438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0097s] [ 40%] 2024-08-06T21:15:51.6626760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 40%] 2024-08-06T21:15:51.6628042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 40%] 2024-08-06T21:15:51.6629282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0071s] [ 40%] 2024-08-06T21:15:51.6630502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 40%] 2024-08-06T21:15:51.6631721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 40%] 2024-08-06T21:15:51.6632992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 40%] 2024-08-06T21:15:51.6634216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0095s] [ 40%] 2024-08-06T21:15:51.6635458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0098s] [ 40%] 2024-08-06T21:15:51.6636724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 40%] 2024-08-06T21:15:51.6638018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 40%] 2024-08-06T21:15:51.6639229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0072s] [ 40%] 2024-08-06T21:15:51.6640460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 40%] 2024-08-06T21:15:51.6641676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 40%] 2024-08-06T21:15:51.6642921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 40%] 2024-08-06T21:15:51.6644157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0092s] [ 40%] 2024-08-06T21:15:51.6645420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0093s] [ 40%] 2024-08-06T21:15:51.6646711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 40%] 2024-08-06T21:15:51.6647937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 40%] 2024-08-06T21:15:51.6649174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0106s] [ 40%] 2024-08-06T21:15:51.6650407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0115s] [ 40%] 2024-08-06T21:15:51.6651652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0062s] [ 40%] 2024-08-06T21:15:51.6652900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 40%] 2024-08-06T21:15:51.6654123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0151s] [ 40%] 2024-08-06T21:15:51.6655417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0152s] [ 40%] 2024-08-06T21:15:51.6656694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0073s] [ 40%] 2024-08-06T21:15:51.6657944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 40%] 2024-08-06T21:15:51.6659155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0080s] [ 40%] 2024-08-06T21:15:51.6660392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 40%] 2024-08-06T21:15:51.6661612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0065s] [ 40%] 2024-08-06T21:15:51.6662866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 40%] 2024-08-06T21:15:51.6664135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0098s] [ 40%] 2024-08-06T21:15:51.6665429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0097s] [ 40%] 2024-08-06T21:15:51.6666670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 40%] 2024-08-06T21:15:51.6667897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 40%] 2024-08-06T21:15:51.6669138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0109s] [ 40%] 2024-08-06T21:15:51.6670369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0121s] [ 40%] 2024-08-06T21:15:51.6671602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0065s] [ 40%] 2024-08-06T21:15:51.6672848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 41%] 2024-08-06T21:15:51.6674137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0150s] [ 41%] 2024-08-06T21:15:51.6675422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0154s] [ 41%] 2024-08-06T21:15:51.6676644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0072s] [ 41%] 2024-08-06T21:15:51.6677897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 41%] 2024-08-06T21:15:51.6679117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0085s] [ 41%] 2024-08-06T21:15:51.6680378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0085s] [ 41%] 2024-08-06T21:15:51.6681589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0065s] [ 41%] 2024-08-06T21:15:51.6682873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 41%] 2024-08-06T21:15:51.6684147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0100s] [ 41%] 2024-08-06T21:15:51.6685394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0102s] [ 41%] 2024-08-06T21:15:51.6686612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0073s] [ 41%] 2024-08-06T21:15:51.6687840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 41%] 2024-08-06T21:15:51.6689090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0125s] [ 41%] 2024-08-06T21:15:51.6690329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0135s] [ 41%] 2024-08-06T21:15:51.6691566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0066s] [ 41%] 2024-08-06T21:15:51.6692832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 41%] 2024-08-06T21:15:51.6694132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0160s] [ 41%] 2024-08-06T21:15:51.6695365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0160s] [ 41%] 2024-08-06T21:15:51.6696605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0073s] [ 41%] 2024-08-06T21:15:51.6697843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 41%] 2024-08-06T21:15:51.6699059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0089s] [ 41%] 2024-08-06T21:15:51.6700311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0094s] [ 41%] 2024-08-06T21:15:51.6701572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0069s] [ 41%] 2024-08-06T21:15:51.6702863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 41%] 2024-08-06T21:15:51.6704497Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0107s] [ 41%] 2024-08-06T21:15:51.6705771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 41%] 2024-08-06T21:15:51.6706995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0075s] [ 41%] 2024-08-06T21:15:51.6708248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 41%] 2024-08-06T21:15:51.6709464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0107s] [ 41%] 2024-08-06T21:15:51.6710706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0113s] [ 41%] 2024-08-06T21:15:51.6712032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0062s] [ 41%] 2024-08-06T21:15:51.6713333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 41%] 2024-08-06T21:15:51.6714575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0144s] [ 41%] 2024-08-06T21:15:51.6715811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0149s] [ 41%] 2024-08-06T21:15:51.6717060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0072s] [ 41%] 2024-08-06T21:15:51.6718288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 41%] 2024-08-06T21:15:51.6719516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0078s] [ 41%] 2024-08-06T21:15:51.6720816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 41%] 2024-08-06T21:15:51.6722103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 41%] 2024-08-06T21:15:51.6723340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 41%] 2024-08-06T21:15:51.6724554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0096s] [ 41%] 2024-08-06T21:15:51.6725801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0099s] [ 41%] 2024-08-06T21:15:51.6727021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 41%] 2024-08-06T21:15:51.6728255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 41%] 2024-08-06T21:15:51.6729473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 41%] 2024-08-06T21:15:51.6730776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 41%] 2024-08-06T21:15:51.6732042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 41%] 2024-08-06T21:15:51.6733258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 41%] 2024-08-06T21:15:51.6734496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 41%] 2024-08-06T21:15:51.6735723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 41%] 2024-08-06T21:15:51.6736957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 41%] 2024-08-06T21:15:51.6738175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 41%] 2024-08-06T21:15:51.6739460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 41%] 2024-08-06T21:15:51.6740754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 41%] 2024-08-06T21:15:51.6741987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 41%] 2024-08-06T21:15:51.6743196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 41%] 2024-08-06T21:15:51.6744416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 41%] 2024-08-06T21:15:51.6745654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 41%] 2024-08-06T21:15:51.6746861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 41%] 2024-08-06T21:15:51.6748092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 41%] 2024-08-06T21:15:51.6749348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 41%] 2024-08-06T21:15:51.6750660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 41%] 2024-08-06T21:15:51.6751874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 41%] 2024-08-06T21:15:51.6753113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 41%] 2024-08-06T21:15:51.6754335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 41%] 2024-08-06T21:15:51.6755583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 41%] 2024-08-06T21:15:51.6756798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 41%] 2024-08-06T21:15:51.6758067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 41%] 2024-08-06T21:15:51.6759348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0062s] [ 41%] 2024-08-06T21:15:51.6760589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 41%] 2024-08-06T21:15:51.6761809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 41%] 2024-08-06T21:15:51.6763028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 41%] 2024-08-06T21:15:51.6764263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 41%] 2024-08-06T21:15:51.6765480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 41%] 2024-08-06T21:15:51.6766711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 41%] 2024-08-06T21:15:51.6767973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 41%] 2024-08-06T21:15:51.6769241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0062s] [ 42%] 2024-08-06T21:15:51.6770502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 42%] 2024-08-06T21:15:51.6771714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 42%] 2024-08-06T21:15:51.6772949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 42%] 2024-08-06T21:15:51.6774177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0076s] [ 42%] 2024-08-06T21:15:51.6775414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 42%] 2024-08-06T21:15:51.6776675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 42%] 2024-08-06T21:15:51.6777978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 42%] 2024-08-06T21:15:51.6779191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0066s] [ 42%] 2024-08-06T21:15:51.6780403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 42%] 2024-08-06T21:15:51.6781633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 42%] 2024-08-06T21:15:51.6782851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 42%] 2024-08-06T21:15:51.6784092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0077s] [ 42%] 2024-08-06T21:15:51.6785309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0078s] [ 42%] 2024-08-06T21:15:51.6786577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 42%] 2024-08-06T21:15:51.6787853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 42%] 2024-08-06T21:15:51.6789090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 42%] 2024-08-06T21:15:51.6790362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 42%] 2024-08-06T21:15:51.6791578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 42%] 2024-08-06T21:15:51.6792819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 42%] 2024-08-06T21:15:51.6794029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0068s] [ 42%] 2024-08-06T21:15:51.6795317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 42%] 2024-08-06T21:15:51.6796591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 42%] 2024-08-06T21:15:51.6797840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 42%] 2024-08-06T21:15:51.6799056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 42%] 2024-08-06T21:15:51.6800292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 42%] 2024-08-06T21:15:51.6801501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 42%] 2024-08-06T21:15:51.6802704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 42%] 2024-08-06T21:15:51.6804290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0072s] [ 42%] 2024-08-06T21:15:51.6805649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 42%] 2024-08-06T21:15:51.6806958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 42%] 2024-08-06T21:15:51.6808168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 42%] 2024-08-06T21:15:51.6809419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0177s] [ 42%] 2024-08-06T21:15:51.6810655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0190s] [ 42%] 2024-08-06T21:15:51.6811905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0077s] [ 42%] 2024-08-06T21:15:51.6813133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0085s] [ 42%] 2024-08-06T21:15:51.6814433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0249s] [ 42%] 2024-08-06T21:15:51.6815759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0257s] [ 42%] 2024-08-06T21:15:51.6816985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0084s] [ 42%] 2024-08-06T21:15:51.6818234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 42%] 2024-08-06T21:15:51.6819467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0094s] [ 42%] 2024-08-06T21:15:51.6820714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0093s] [ 42%] 2024-08-06T21:15:51.6821923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0074s] [ 42%] 2024-08-06T21:15:51.6823162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 42%] 2024-08-06T21:15:51.6824432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0113s] [ 42%] 2024-08-06T21:15:51.6825711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0111s] [ 42%] 2024-08-06T21:15:51.6826951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0080s] [ 42%] 2024-08-06T21:15:51.6828170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 42%] 2024-08-06T21:15:51.6829428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0182s] [ 42%] 2024-08-06T21:15:51.6830658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0199s] [ 42%] 2024-08-06T21:15:51.6831893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0078s] [ 42%] 2024-08-06T21:15:51.6833163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0081s] [ 42%] 2024-08-06T21:15:51.6834454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0255s] [ 42%] 2024-08-06T21:15:51.6835687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0261s] [ 42%] 2024-08-06T21:15:51.6836910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0087s] [ 42%] 2024-08-06T21:15:51.6838163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0087s] [ 42%] 2024-08-06T21:15:51.6839403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0098s] [ 42%] 2024-08-06T21:15:51.6840640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0097s] [ 42%] 2024-08-06T21:15:51.6841847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0077s] [ 42%] 2024-08-06T21:15:51.6843128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0076s] [ 42%] 2024-08-06T21:15:51.6844403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0116s] [ 42%] 2024-08-06T21:15:51.6845639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0113s] [ 42%] 2024-08-06T21:15:51.6846857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0082s] [ 42%] 2024-08-06T21:15:51.6848083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0081s] [ 42%] 2024-08-06T21:15:51.6849344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0201s] [ 42%] 2024-08-06T21:15:51.6850572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0222s] [ 42%] 2024-08-06T21:15:51.6851852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0089s] [ 42%] 2024-08-06T21:15:51.6853135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 42%] 2024-08-06T21:15:51.6854377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0271s] [ 42%] 2024-08-06T21:15:51.6855601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0277s] [ 42%] 2024-08-06T21:15:51.6856849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0093s] [ 42%] 2024-08-06T21:15:51.6858082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0096s] [ 42%] 2024-08-06T21:15:51.6859306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0107s] [ 42%] 2024-08-06T21:15:51.6860538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0107s] [ 42%] 2024-08-06T21:15:51.6861799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0079s] [ 42%] 2024-08-06T21:15:51.6863093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 42%] 2024-08-06T21:15:51.6864311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0125s] [ 42%] 2024-08-06T21:15:51.6865553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0126s] [ 42%] 2024-08-06T21:15:51.6866771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0088s] [ 43%] 2024-08-06T21:15:51.6868020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0085s] [ 43%] 2024-08-06T21:15:51.6869249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0169s] [ 43%] 2024-08-06T21:15:51.6870514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0182s] [ 43%] 2024-08-06T21:15:51.6871801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0076s] [ 43%] 2024-08-06T21:15:51.6873027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 43%] 2024-08-06T21:15:51.6874260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0244s] [ 43%] 2024-08-06T21:15:51.6875485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0253s] [ 43%] 2024-08-06T21:15:51.6876728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0084s] [ 43%] 2024-08-06T21:15:51.6877954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 43%] 2024-08-06T21:15:51.6879191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0089s] [ 43%] 2024-08-06T21:15:51.6880463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0092s] [ 43%] 2024-08-06T21:15:51.6881733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 43%] 2024-08-06T21:15:51.6882969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 43%] 2024-08-06T21:15:51.6884181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0110s] [ 43%] 2024-08-06T21:15:51.6885421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0109s] [ 43%] 2024-08-06T21:15:51.6886643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0080s] [ 43%] 2024-08-06T21:15:51.6887877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 43%] 2024-08-06T21:15:51.6889148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 43%] 2024-08-06T21:15:51.6890441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 43%] 2024-08-06T21:15:51.6891660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 43%] 2024-08-06T21:15:51.6892883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 43%] 2024-08-06T21:15:51.6894132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 43%] 2024-08-06T21:15:51.6895366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 43%] 2024-08-06T21:15:51.6896602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 43%] 2024-08-06T21:15:51.6897825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 43%] 2024-08-06T21:15:51.6899094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 43%] 2024-08-06T21:15:51.6900389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 43%] 2024-08-06T21:15:51.6901614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 43%] 2024-08-06T21:15:51.6902828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 43%] 2024-08-06T21:15:51.6904439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 43%] 2024-08-06T21:15:51.6905720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 43%] 2024-08-06T21:15:51.6906936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 43%] 2024-08-06T21:15:51.6908248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 43%] 2024-08-06T21:15:51.6909541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 43%] 2024-08-06T21:15:51.6910809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 43%] 2024-08-06T21:15:51.6912024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 43%] 2024-08-06T21:15:51.6913268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 43%] 2024-08-06T21:15:51.6914500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0076s] [ 43%] 2024-08-06T21:15:51.6915720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 43%] 2024-08-06T21:15:51.6916957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 43%] 2024-08-06T21:15:51.6918244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 43%] 2024-08-06T21:15:51.6919594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 43%] 2024-08-06T21:15:51.6920831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 43%] 2024-08-06T21:15:51.6922059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 43%] 2024-08-06T21:15:51.6923281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 43%] 2024-08-06T21:15:51.6924524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 43%] 2024-08-06T21:15:51.6925741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 43%] 2024-08-06T21:15:51.6927001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 43%] 2024-08-06T21:15:51.6928295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 43%] 2024-08-06T21:15:51.6929513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 43%] 2024-08-06T21:15:51.6930778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 43%] 2024-08-06T21:15:51.6931998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 43%] 2024-08-06T21:15:51.6933241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 43%] 2024-08-06T21:15:51.6934459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0077s] [ 43%] 2024-08-06T21:15:51.6935700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0078s] [ 43%] 2024-08-06T21:15:51.6936964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 43%] 2024-08-06T21:15:51.6938905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 43%] 2024-08-06T21:15:51.6940136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0064s] [ 43%] 2024-08-06T21:15:51.6941374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 43%] 2024-08-06T21:15:51.6942607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 43%] 2024-08-06T21:15:51.6943829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 43%] 2024-08-06T21:15:51.6945055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0081s] [ 43%] 2024-08-06T21:15:51.6946317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0081s] [ 43%] 2024-08-06T21:15:51.6947614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 43%] 2024-08-06T21:15:51.6948839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 43%] 2024-08-06T21:15:51.6950050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 43%] 2024-08-06T21:15:51.6951316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 43%] 2024-08-06T21:15:51.6952538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 43%] 2024-08-06T21:15:51.6953772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 43%] 2024-08-06T21:15:51.6954988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0075s] [ 43%] 2024-08-06T21:15:51.6956275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 43%] 2024-08-06T21:15:51.6957546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 43%] 2024-08-06T21:15:51.6958790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 43%] 2024-08-06T21:15:51.6959993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 43%] 2024-08-06T21:15:51.6961234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 43%] 2024-08-06T21:15:51.6962461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 43%] 2024-08-06T21:15:51.6963672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 44%] 2024-08-06T21:15:51.6964946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0077s] [ 44%] 2024-08-06T21:15:51.6966215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 44%] 2024-08-06T21:15:51.6967448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 44%] 2024-08-06T21:15:51.6968660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 44%] 2024-08-06T21:15:51.6969891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 44%] 2024-08-06T21:15:51.6971133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 44%] 2024-08-06T21:15:51.6972363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 44%] 2024-08-06T21:15:51.6973581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 44%] 2024-08-06T21:15:51.6974840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0071s] [ 44%] 2024-08-06T21:15:51.6976132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 44%] 2024-08-06T21:15:51.6977348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 44%] 2024-08-06T21:15:51.6978589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 44%] 2024-08-06T21:15:51.6979796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 44%] 2024-08-06T21:15:51.6981054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 44%] 2024-08-06T21:15:51.6982260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 44%] 2024-08-06T21:15:51.6983532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 44%] 2024-08-06T21:15:51.6984793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0073s] [ 44%] 2024-08-06T21:15:51.6986014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 44%] 2024-08-06T21:15:51.6987239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 44%] 2024-08-06T21:15:51.6988454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 44%] 2024-08-06T21:15:51.6989687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 44%] 2024-08-06T21:15:51.6990926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 44%] 2024-08-06T21:15:51.6992150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 44%] 2024-08-06T21:15:51.6993417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 44%] 2024-08-06T21:15:51.6994713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0073s] [ 44%] 2024-08-06T21:15:51.6995962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 44%] 2024-08-06T21:15:51.6997187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 44%] 2024-08-06T21:15:51.6998433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 44%] 2024-08-06T21:15:51.6999647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 44%] 2024-08-06T21:15:51.7000872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 44%] 2024-08-06T21:15:51.7002118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 44%] 2024-08-06T21:15:51.7003447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 44%] 2024-08-06T21:15:51.7005069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0075s] [ 44%] 2024-08-06T21:15:51.7006342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 44%] 2024-08-06T21:15:51.7007558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 44%] 2024-08-06T21:15:51.7008787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 44%] 2024-08-06T21:15:51.7010020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 44%] 2024-08-06T21:15:51.7011233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 44%] 2024-08-06T21:15:51.7012538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 44%] 2024-08-06T21:15:51.7013835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 44%] 2024-08-06T21:15:51.7015065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0077s] [ 44%] 2024-08-06T21:15:51.7016299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 44%] 2024-08-06T21:15:51.7017541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 44%] 2024-08-06T21:15:51.7018769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 44%] 2024-08-06T21:15:51.7019972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 44%] 2024-08-06T21:15:51.7021204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 44%] 2024-08-06T21:15:51.7022483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 44%] 2024-08-06T21:15:51.7023781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 44%] 2024-08-06T21:15:51.7024993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0077s] [ 44%] 2024-08-06T21:15:51.7026239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 44%] 2024-08-06T21:15:51.7027454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 44%] 2024-08-06T21:15:51.7028691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 44%] 2024-08-06T21:15:51.7029896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 44%] 2024-08-06T21:15:51.7031154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 44%] 2024-08-06T21:15:51.7032435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 44%] 2024-08-06T21:15:51.7033650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 44%] 2024-08-06T21:15:51.7034877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0069s] [ 44%] 2024-08-06T21:15:51.7036108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 44%] 2024-08-06T21:15:51.7037347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 44%] 2024-08-06T21:15:51.7038564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 44%] 2024-08-06T21:15:51.7039783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 44%] 2024-08-06T21:15:51.7041038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 44%] 2024-08-06T21:15:51.7042293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 44%] 2024-08-06T21:15:51.7043519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 44%] 2024-08-06T21:15:51.7044731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0072s] [ 44%] 2024-08-06T21:15:51.7045967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 44%] 2024-08-06T21:15:51.7047174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 44%] 2024-08-06T21:15:51.7048400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 44%] 2024-08-06T21:15:51.7049620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0101s] [ 44%] 2024-08-06T21:15:51.7050908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 44%] 2024-08-06T21:15:51.7052176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 44%] 2024-08-06T21:15:51.7053398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 44%] 2024-08-06T21:15:51.7054645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0140s] [ 44%] 2024-08-06T21:15:51.7055885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0141s] [ 44%] 2024-08-06T21:15:51.7057123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 44%] 2024-08-06T21:15:51.7058352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 44%] 2024-08-06T21:15:51.7059622Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0096s] [ 45%] 2024-08-06T21:15:51.7060891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0100s] [ 45%] 2024-08-06T21:15:51.7062124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 45%] 2024-08-06T21:15:51.7063346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 45%] 2024-08-06T21:15:51.7064572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0134s] [ 45%] 2024-08-06T21:15:51.7065823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0137s] [ 45%] 2024-08-06T21:15:51.7067040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 45%] 2024-08-06T21:15:51.7068328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 45%] 2024-08-06T21:15:51.7069600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0104s] [ 45%] 2024-08-06T21:15:51.7070841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0111s] [ 45%] 2024-08-06T21:15:51.7072058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0062s] [ 45%] 2024-08-06T21:15:51.7073303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 45%] 2024-08-06T21:15:51.7074533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0143s] [ 45%] 2024-08-06T21:15:51.7075763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0143s] [ 45%] 2024-08-06T21:15:51.7076997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 45%] 2024-08-06T21:15:51.7078273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 45%] 2024-08-06T21:15:51.7079562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0101s] [ 45%] 2024-08-06T21:15:51.7080802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 45%] 2024-08-06T21:15:51.7082035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 45%] 2024-08-06T21:15:51.7083256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 45%] 2024-08-06T21:15:51.7084496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0137s] [ 45%] 2024-08-06T21:15:51.7085719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0140s] [ 45%] 2024-08-06T21:15:51.7086984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0073s] [ 45%] 2024-08-06T21:15:51.7088291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 45%] 2024-08-06T21:15:51.7089561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0113s] [ 45%] 2024-08-06T21:15:51.7090827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0125s] [ 45%] 2024-08-06T21:15:51.7092054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0065s] [ 45%] 2024-08-06T21:15:51.7093305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 45%] 2024-08-06T21:15:51.7094531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0152s] [ 45%] 2024-08-06T21:15:51.7095774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0150s] [ 45%] 2024-08-06T21:15:51.7097040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 45%] 2024-08-06T21:15:51.7098323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 45%] 2024-08-06T21:15:51.7099553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0110s] [ 45%] 2024-08-06T21:15:51.7100789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0116s] [ 45%] 2024-08-06T21:15:51.7102026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0066s] [ 45%] 2024-08-06T21:15:51.7103285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 45%] 2024-08-06T21:15:51.7104926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0142s] [ 45%] 2024-08-06T21:15:51.7106262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0146s] [ 45%] 2024-08-06T21:15:51.7107578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0072s] [ 45%] 2024-08-06T21:15:51.7108803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 45%] 2024-08-06T21:15:51.7110022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0099s] [ 45%] 2024-08-06T21:15:51.7111284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0101s] [ 45%] 2024-08-06T21:15:51.7112511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 45%] 2024-08-06T21:15:51.7113755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 45%] 2024-08-06T21:15:51.7114977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0142s] [ 45%] 2024-08-06T21:15:51.7116291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0142s] [ 45%] 2024-08-06T21:15:51.7117592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 45%] 2024-08-06T21:15:51.7118833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 45%] 2024-08-06T21:15:51.7120050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0095s] [ 45%] 2024-08-06T21:15:51.7121295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0099s] [ 45%] 2024-08-06T21:15:51.7122532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 45%] 2024-08-06T21:15:51.7123746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 45%] 2024-08-06T21:15:51.7125028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0134s] [ 45%] 2024-08-06T21:15:51.7126305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0136s] [ 45%] 2024-08-06T21:15:51.7127540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 45%] 2024-08-06T21:15:51.7128759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 45%] 2024-08-06T21:15:51.7130002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0155s] [ 45%] 2024-08-06T21:15:51.7131257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0168s] [ 45%] 2024-08-06T21:15:51.7132478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0071s] [ 45%] 2024-08-06T21:15:51.7133762Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 45%] 2024-08-06T21:15:51.7135040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0233s] [ 45%] 2024-08-06T21:15:51.7136295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0238s] [ 45%] 2024-08-06T21:15:51.7137516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0081s] [ 45%] 2024-08-06T21:15:51.7138765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 45%] 2024-08-06T21:15:51.7139987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0128s] [ 45%] 2024-08-06T21:15:51.7141248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0132s] [ 45%] 2024-08-06T21:15:51.7142459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 45%] 2024-08-06T21:15:51.7143723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 45%] 2024-08-06T21:15:51.7145019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0180s] [ 45%] 2024-08-06T21:15:51.7146243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0183s] [ 45%] 2024-08-06T21:15:51.7147476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0079s] [ 45%] 2024-08-06T21:15:51.7148704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0076s] [ 45%] 2024-08-06T21:15:51.7149952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0163s] [ 45%] 2024-08-06T21:15:51.7151192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0179s] [ 45%] 2024-08-06T21:15:51.7152477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0072s] [ 45%] 2024-08-06T21:15:51.7153760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 45%] 2024-08-06T21:15:51.7154988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0237s] [ 45%] 2024-08-06T21:15:51.7156234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0240s] [ 45%] 2024-08-06T21:15:51.7157488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0082s] [ 46%] 2024-08-06T21:15:51.7158745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 46%] 2024-08-06T21:15:51.7159965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0132s] [ 46%] 2024-08-06T21:15:51.7161199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0139s] [ 46%] 2024-08-06T21:15:51.7162471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0072s] [ 46%] 2024-08-06T21:15:51.7163769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 46%] 2024-08-06T21:15:51.7164986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0182s] [ 46%] 2024-08-06T21:15:51.7166210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0186s] [ 46%] 2024-08-06T21:15:51.7167475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0079s] [ 46%] 2024-08-06T21:15:51.7168712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 46%] 2024-08-06T21:15:51.7169954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0179s] [ 46%] 2024-08-06T21:15:51.7171225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0200s] [ 46%] 2024-08-06T21:15:51.7172520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0079s] [ 46%] 2024-08-06T21:15:51.7173748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 46%] 2024-08-06T21:15:51.7174991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0251s] [ 46%] 2024-08-06T21:15:51.7176225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0255s] [ 46%] 2024-08-06T21:15:51.7177498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0089s] [ 46%] 2024-08-06T21:15:51.7178733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0089s] [ 46%] 2024-08-06T21:15:51.7179944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0147s] [ 46%] 2024-08-06T21:15:51.7181225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0154s] [ 46%] 2024-08-06T21:15:51.7182493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0076s] [ 46%] 2024-08-06T21:15:51.7183733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0076s] [ 46%] 2024-08-06T21:15:51.7184948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0198s] [ 46%] 2024-08-06T21:15:51.7186197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0198s] [ 46%] 2024-08-06T21:15:51.7187443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0084s] [ 46%] 2024-08-06T21:15:51.7188690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0083s] [ 46%] 2024-08-06T21:15:51.7189949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0149s] [ 46%] 2024-08-06T21:15:51.7191227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0161s] [ 46%] 2024-08-06T21:15:51.7192471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0071s] [ 46%] 2024-08-06T21:15:51.7193692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 46%] 2024-08-06T21:15:51.7194933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0228s] [ 46%] 2024-08-06T21:15:51.7196163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0232s] [ 46%] 2024-08-06T21:15:51.7197425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0077s] [ 46%] 2024-08-06T21:15:51.7198650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 46%] 2024-08-06T21:15:51.7199930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0123s] [ 46%] 2024-08-06T21:15:51.7201204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0128s] [ 46%] 2024-08-06T21:15:51.7202414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0068s] [ 46%] 2024-08-06T21:15:51.7203688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 46%] 2024-08-06T21:15:51.7205279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0177s] [ 46%] 2024-08-06T21:15:51.7206549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0177s] [ 46%] 2024-08-06T21:15:51.7207785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0076s] [ 46%] 2024-08-06T21:15:51.7209101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0077s] [ 46%] 2024-08-06T21:15:51.7210389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0071s] [ 46%] 2024-08-06T21:15:51.7211630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0070s] [ 46%] 2024-08-06T21:15:51.7212842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 46%] 2024-08-06T21:15:51.7214064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 46%] 2024-08-06T21:15:51.7215302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0093s] [ 46%] 2024-08-06T21:15:51.7216521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0093s] [ 46%] 2024-08-06T21:15:51.7217773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 46%] 2024-08-06T21:15:51.7219102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 46%] 2024-08-06T21:15:51.7220399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0072s] [ 46%] 2024-08-06T21:15:51.7221610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 46%] 2024-08-06T21:15:51.7222831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 46%] 2024-08-06T21:15:51.7224042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 46%] 2024-08-06T21:15:51.7225257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0096s] [ 46%] 2024-08-06T21:15:51.7226496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0095s] [ 46%] 2024-08-06T21:15:51.7227767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 46%] 2024-08-06T21:15:51.7229072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 46%] 2024-08-06T21:15:51.7230287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0073s] [ 46%] 2024-08-06T21:15:51.7231519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 46%] 2024-08-06T21:15:51.7232733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 46%] 2024-08-06T21:15:51.7233974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 46%] 2024-08-06T21:15:51.7235193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0094s] [ 46%] 2024-08-06T21:15:51.7236442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0095s] [ 46%] 2024-08-06T21:15:51.7237726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 46%] 2024-08-06T21:15:51.7239003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 46%] 2024-08-06T21:15:51.7240229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0078s] [ 46%] 2024-08-06T21:15:51.7241443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 46%] 2024-08-06T21:15:51.7242668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 46%] 2024-08-06T21:15:51.7243887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 46%] 2024-08-06T21:15:51.7245119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0099s] [ 46%] 2024-08-06T21:15:51.7246392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0099s] [ 46%] 2024-08-06T21:15:51.7247656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 46%] 2024-08-06T21:15:51.7248893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 46%] 2024-08-06T21:15:51.7250106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0079s] [ 46%] 2024-08-06T21:15:51.7251343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 46%] 2024-08-06T21:15:51.7252566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 46%] 2024-08-06T21:15:51.7253803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 47%] 2024-08-06T21:15:51.7255016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0102s] [ 47%] 2024-08-06T21:15:51.7256309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 47%] 2024-08-06T21:15:51.7257579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 47%] 2024-08-06T21:15:51.7258800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 47%] 2024-08-06T21:15:51.7260024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0082s] [ 47%] 2024-08-06T21:15:51.7261240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0083s] [ 47%] 2024-08-06T21:15:51.7262471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0062s] [ 47%] 2024-08-06T21:15:51.7263680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 47%] 2024-08-06T21:15:51.7264904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0105s] [ 47%] 2024-08-06T21:15:51.7266175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0104s] [ 47%] 2024-08-06T21:15:51.7267458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 47%] 2024-08-06T21:15:51.7268675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 47%] 2024-08-06T21:15:51.7269889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0068s] [ 47%] 2024-08-06T21:15:51.7271124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0070s] [ 47%] 2024-08-06T21:15:51.7272334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 47%] 2024-08-06T21:15:51.7273561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 47%] 2024-08-06T21:15:51.7274819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0093s] [ 47%] 2024-08-06T21:15:51.7276114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0091s] [ 47%] 2024-08-06T21:15:51.7277323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 47%] 2024-08-06T21:15:51.7278553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 47%] 2024-08-06T21:15:51.7279761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0071s] [ 47%] 2024-08-06T21:15:51.7280977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 47%] 2024-08-06T21:15:51.7282193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 47%] 2024-08-06T21:15:51.7283397Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 47%] 2024-08-06T21:15:51.7284666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0092s] [ 47%] 2024-08-06T21:15:51.7285925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0094s] [ 47%] 2024-08-06T21:15:51.7287155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 47%] 2024-08-06T21:15:51.7288364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 47%] 2024-08-06T21:15:51.7289613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0279s] [ 47%] 2024-08-06T21:15:51.7290843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0306s] [ 47%] 2024-08-06T21:15:51.7292065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0095s] [ 47%] 2024-08-06T21:15:51.7293349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0095s] [ 47%] 2024-08-06T21:15:51.7294629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0419s] [ 47%] 2024-08-06T21:15:51.7295876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0436s] [ 47%] 2024-08-06T21:15:51.7297099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0109s] [ 47%] 2024-08-06T21:15:51.7298354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0110s] [ 47%] 2024-08-06T21:15:51.7299571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0149s] [ 47%] 2024-08-06T21:15:51.7300806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0154s] [ 47%] 2024-08-06T21:15:51.7302018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0088s] [ 47%] 2024-08-06T21:15:51.7303328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0090s] [ 47%] 2024-08-06T21:15:51.7305003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0203s] [ 47%] 2024-08-06T21:15:51.7306269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0206s] [ 47%] 2024-08-06T21:15:51.7307508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0100s] [ 47%] 2024-08-06T21:15:51.7308740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0102s] [ 47%] 2024-08-06T21:15:51.7309978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0289s] [ 47%] 2024-08-06T21:15:51.7311202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0322s] [ 47%] 2024-08-06T21:15:51.7312534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0100s] [ 47%] 2024-08-06T21:15:51.7313842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0102s] [ 47%] 2024-08-06T21:15:51.7315069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0425s] [ 47%] 2024-08-06T21:15:51.7316331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0439s] [ 47%] 2024-08-06T21:15:51.7317560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0114s] [ 47%] 2024-08-06T21:15:51.7318818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0112s] [ 47%] 2024-08-06T21:15:51.7320032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0155s] [ 47%] 2024-08-06T21:15:51.7321270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0161s] [ 47%] 2024-08-06T21:15:51.7322540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0093s] [ 47%] 2024-08-06T21:15:51.7323842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0094s] [ 47%] 2024-08-06T21:15:51.7325065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0208s] [ 47%] 2024-08-06T21:15:51.7326313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0210s] [ 47%] 2024-08-06T21:15:51.7327561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0103s] [ 47%] 2024-08-06T21:15:51.7328786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0105s] [ 47%] 2024-08-06T21:15:51.7330021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0324s] [ 47%] 2024-08-06T21:15:51.7331290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0360s] [ 47%] 2024-08-06T21:15:51.7332580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0111s] [ 47%] 2024-08-06T21:15:51.7333808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0107s] [ 47%] 2024-08-06T21:15:51.7335047Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0455s] [ 47%] 2024-08-06T21:15:51.7336305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0467s] [ 47%] 2024-08-06T21:15:51.7337539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0121s] [ 47%] 2024-08-06T21:15:51.7338786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0121s] [ 47%] 2024-08-06T21:15:51.7340000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0173s] [ 47%] 2024-08-06T21:15:51.7341283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0181s] [ 47%] 2024-08-06T21:15:51.7342558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0102s] [ 47%] 2024-08-06T21:15:51.7343793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0102s] [ 47%] 2024-08-06T21:15:51.7345015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0225s] [ 47%] 2024-08-06T21:15:51.7346269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0228s] [ 47%] 2024-08-06T21:15:51.7347495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0110s] [ 47%] 2024-08-06T21:15:51.7348714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0113s] [ 47%] 2024-08-06T21:15:51.7349994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0266s] [ 47%] 2024-08-06T21:15:51.7351277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0288s] [ 48%] 2024-08-06T21:15:51.7352512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0093s] [ 48%] 2024-08-06T21:15:51.7353733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0100s] [ 48%] 2024-08-06T21:15:51.7354977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0412s] [ 48%] 2024-08-06T21:15:51.7356219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0429s] [ 48%] 2024-08-06T21:15:51.7357452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0106s] [ 48%] 2024-08-06T21:15:51.7358719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0106s] [ 48%] 2024-08-06T21:15:51.7359991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0146s] [ 48%] 2024-08-06T21:15:51.7361227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0147s] [ 48%] 2024-08-06T21:15:51.7362452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0086s] [ 48%] 2024-08-06T21:15:51.7363689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0087s] [ 48%] 2024-08-06T21:15:51.7364912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0199s] [ 48%] 2024-08-06T21:15:51.7366152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0201s] [ 48%] 2024-08-06T21:15:51.7367369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0097s] [ 48%] 2024-08-06T21:15:51.7368657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0099s] [ 48%] 2024-08-06T21:15:51.7369941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0077s] [ 48%] 2024-08-06T21:15:51.7371161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 48%] 2024-08-06T21:15:51.7372410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 48%] 2024-08-06T21:15:51.7373637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 48%] 2024-08-06T21:15:51.7374887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0098s] [ 48%] 2024-08-06T21:15:51.7376110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0100s] [ 48%] 2024-08-06T21:15:51.7377393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 48%] 2024-08-06T21:15:51.7378674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 48%] 2024-08-06T21:15:51.7379903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0081s] [ 48%] 2024-08-06T21:15:51.7381112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0081s] [ 48%] 2024-08-06T21:15:51.7382348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0062s] [ 48%] 2024-08-06T21:15:51.7383590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 48%] 2024-08-06T21:15:51.7384806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0103s] [ 48%] 2024-08-06T21:15:51.7386048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0104s] [ 48%] 2024-08-06T21:15:51.7387306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 48%] 2024-08-06T21:15:51.7388595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 48%] 2024-08-06T21:15:51.7389855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0075s] [ 48%] 2024-08-06T21:15:51.7391101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 48%] 2024-08-06T21:15:51.7392337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 48%] 2024-08-06T21:15:51.7393561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 48%] 2024-08-06T21:15:51.7394799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0102s] [ 48%] 2024-08-06T21:15:51.7396063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0104s] [ 48%] 2024-08-06T21:15:51.7397351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 48%] 2024-08-06T21:15:51.7398582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 48%] 2024-08-06T21:15:51.7399804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0079s] [ 48%] 2024-08-06T21:15:51.7401022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 48%] 2024-08-06T21:15:51.7402272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 48%] 2024-08-06T21:15:51.7403525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 48%] 2024-08-06T21:15:51.7405275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0103s] [ 48%] 2024-08-06T21:15:51.7406617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 48%] 2024-08-06T21:15:51.7407906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 48%] 2024-08-06T21:15:51.7409140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 48%] 2024-08-06T21:15:51.7410357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0083s] [ 48%] 2024-08-06T21:15:51.7411597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 48%] 2024-08-06T21:15:51.7412834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0062s] [ 48%] 2024-08-06T21:15:51.7414071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 48%] 2024-08-06T21:15:51.7415290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0104s] [ 48%] 2024-08-06T21:15:51.7416599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0106s] [ 48%] 2024-08-06T21:15:51.7417887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 48%] 2024-08-06T21:15:51.7419113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 48%] 2024-08-06T21:15:51.7420344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0085s] [ 48%] 2024-08-06T21:15:51.7421572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0087s] [ 48%] 2024-08-06T21:15:51.7422816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 48%] 2024-08-06T21:15:51.7424029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 48%] 2024-08-06T21:15:51.7425308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0106s] [ 48%] 2024-08-06T21:15:51.7426583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0108s] [ 48%] 2024-08-06T21:15:51.7427813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 48%] 2024-08-06T21:15:51.7429032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 48%] 2024-08-06T21:15:51.7430252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0074s] [ 48%] 2024-08-06T21:15:51.7431488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 48%] 2024-08-06T21:15:51.7432723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 48%] 2024-08-06T21:15:51.7433955Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 48%] 2024-08-06T21:15:51.7435220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0099s] [ 48%] 2024-08-06T21:15:51.7436515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0099s] [ 48%] 2024-08-06T21:15:51.7437726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 48%] 2024-08-06T21:15:51.7438971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 48%] 2024-08-06T21:15:51.7440190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0078s] [ 48%] 2024-08-06T21:15:51.7441407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 48%] 2024-08-06T21:15:51.7442651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 48%] 2024-08-06T21:15:51.7443918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 48%] 2024-08-06T21:15:51.7445202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0103s] [ 48%] 2024-08-06T21:15:51.7446424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 48%] 2024-08-06T21:15:51.7447651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 49%] 2024-08-06T21:15:51.7448874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 49%] 2024-08-06T21:15:51.7450113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0072s] [ 49%] 2024-08-06T21:15:51.7451332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 49%] 2024-08-06T21:15:51.7452548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 49%] 2024-08-06T21:15:51.7453830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 49%] 2024-08-06T21:15:51.7455114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0094s] [ 49%] 2024-08-06T21:15:51.7456355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0093s] [ 49%] 2024-08-06T21:15:51.7457573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 49%] 2024-08-06T21:15:51.7458817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 49%] 2024-08-06T21:15:51.7460027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0075s] [ 49%] 2024-08-06T21:15:51.7461256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 49%] 2024-08-06T21:15:51.7462504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 49%] 2024-08-06T21:15:51.7463772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 49%] 2024-08-06T21:15:51.7465019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0097s] [ 49%] 2024-08-06T21:15:51.7466238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0098s] [ 49%] 2024-08-06T21:15:51.7467469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 49%] 2024-08-06T21:15:51.7468688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 49%] 2024-08-06T21:15:51.7469915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0073s] [ 49%] 2024-08-06T21:15:51.7471129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 49%] 2024-08-06T21:15:51.7472406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 49%] 2024-08-06T21:15:51.7473676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 49%] 2024-08-06T21:15:51.7474909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0097s] [ 49%] 2024-08-06T21:15:51.7476153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0097s] [ 49%] 2024-08-06T21:15:51.7477374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 49%] 2024-08-06T21:15:51.7478615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 49%] 2024-08-06T21:15:51.7479817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0077s] [ 49%] 2024-08-06T21:15:51.7481050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 49%] 2024-08-06T21:15:51.7482301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 49%] 2024-08-06T21:15:51.7483583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 49%] 2024-08-06T21:15:51.7484813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0099s] [ 49%] 2024-08-06T21:15:51.7486038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0098s] [ 49%] 2024-08-06T21:15:51.7487274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 49%] 2024-08-06T21:15:51.7488489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 49%] 2024-08-06T21:15:51.7489721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0080s] [ 49%] 2024-08-06T21:15:51.7490982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 49%] 2024-08-06T21:15:51.7492258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 49%] 2024-08-06T21:15:51.7493475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 49%] 2024-08-06T21:15:51.7494727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0102s] [ 49%] 2024-08-06T21:15:51.7495958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0102s] [ 49%] 2024-08-06T21:15:51.7497175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 49%] 2024-08-06T21:15:51.7498412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 49%] 2024-08-06T21:15:51.7499618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0083s] [ 49%] 2024-08-06T21:15:51.7509714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 49%] 2024-08-06T21:15:51.7511130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 49%] 2024-08-06T21:15:51.7512359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 49%] 2024-08-06T21:15:51.7513594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0106s] [ 49%] 2024-08-06T21:15:51.7514829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0107s] [ 49%] 2024-08-06T21:15:51.7516033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 49%] 2024-08-06T21:15:51.7517244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 49%] 2024-08-06T21:15:51.7518513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0070s] [ 49%] 2024-08-06T21:15:51.7519856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 49%] 2024-08-06T21:15:51.7521060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 49%] 2024-08-06T21:15:51.7522270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 49%] 2024-08-06T21:15:51.7523519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0093s] [ 49%] 2024-08-06T21:15:51.7524741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0094s] [ 49%] 2024-08-06T21:15:51.7525950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 49%] 2024-08-06T21:15:51.7527164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 49%] 2024-08-06T21:15:51.7528413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0072s] [ 49%] 2024-08-06T21:15:51.7529667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 49%] 2024-08-06T21:15:51.7530860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 49%] 2024-08-06T21:15:51.7532075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 49%] 2024-08-06T21:15:51.7533307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0095s] [ 49%] 2024-08-06T21:15:51.7534519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0096s] [ 49%] 2024-08-06T21:15:51.7535712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 49%] 2024-08-06T21:15:51.7536971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 49%] 2024-08-06T21:15:51.7538224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 49%] 2024-08-06T21:15:51.7539436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0050s] [ 49%] 2024-08-06T21:15:51.7540629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 49%] 2024-08-06T21:15:51.7541844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 49%] 2024-08-06T21:15:51.7543088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 49%] 2024-08-06T21:15:51.7544301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 49%] 2024-08-06T21:15:51.7545515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 49%] 2024-08-06T21:15:51.7546771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 49%] 2024-08-06T21:15:51.7548026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 49%] 2024-08-06T21:15:51.7549224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 49%] 2024-08-06T21:15:51.7550426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 49%] 2024-08-06T21:15:51.7551631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 50%] 2024-08-06T21:15:51.7552855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 50%] 2024-08-06T21:15:51.7554067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 50%] 2024-08-06T21:15:51.7555308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 50%] 2024-08-06T21:15:51.7556573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 50%] 2024-08-06T21:15:51.7557770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 50%] 2024-08-06T21:15:51.7558990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 50%] 2024-08-06T21:15:51.7560196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 50%] 2024-08-06T21:15:51.7561417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 50%] 2024-08-06T21:15:51.7562640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 50%] 2024-08-06T21:15:51.7563853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 50%] 2024-08-06T21:15:51.7565110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 50%] 2024-08-06T21:15:51.7566375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 50%] 2024-08-06T21:15:51.7567574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 50%] 2024-08-06T21:15:51.7568772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 50%] 2024-08-06T21:15:51.7569980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 50%] 2024-08-06T21:15:51.7571187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 50%] 2024-08-06T21:15:51.7572417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 50%] 2024-08-06T21:15:51.7573668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 50%] 2024-08-06T21:15:51.7574921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 50%] 2024-08-06T21:15:51.7576132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 50%] 2024-08-06T21:15:51.7577332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 50%] 2024-08-06T21:15:51.7578548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 50%] 2024-08-06T21:15:51.7579752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 50%] 2024-08-06T21:15:51.7580960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0049s] [ 50%] 2024-08-06T21:15:51.7582180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 50%] 2024-08-06T21:15:51.7583444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 50%] 2024-08-06T21:15:51.7584704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 50%] 2024-08-06T21:15:51.7585913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 50%] 2024-08-06T21:15:51.7587122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 50%] 2024-08-06T21:15:51.7588327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 50%] 2024-08-06T21:15:51.7589527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 50%] 2024-08-06T21:15:51.7590727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 50%] 2024-08-06T21:15:51.7592003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 50%] 2024-08-06T21:15:51.7593270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 50%] 2024-08-06T21:15:51.7594473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 50%] 2024-08-06T21:15:51.7595678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 50%] 2024-08-06T21:15:51.7596882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 50%] 2024-08-06T21:15:51.7598102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 50%] 2024-08-06T21:15:51.7599300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 50%] 2024-08-06T21:15:51.7600510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0049s] [ 50%] 2024-08-06T21:15:51.7601763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 50%] 2024-08-06T21:15:51.7603040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 50%] 2024-08-06T21:15:51.7604722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 50%] 2024-08-06T21:15:51.7605981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 50%] 2024-08-06T21:15:51.7607184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 50%] 2024-08-06T21:15:51.7608387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 50%] 2024-08-06T21:15:51.7609584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 50%] 2024-08-06T21:15:51.7610866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 50%] 2024-08-06T21:15:51.7612154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 50%] 2024-08-06T21:15:51.7613360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 50%] 2024-08-06T21:15:51.7614557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 50%] 2024-08-06T21:15:51.7615765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 50%] 2024-08-06T21:15:51.7616986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 50%] 2024-08-06T21:15:51.7618201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 50%] 2024-08-06T21:15:51.7619401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 50%] 2024-08-06T21:15:51.7620688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 50%] 2024-08-06T21:15:51.7621969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 50%] 2024-08-06T21:15:51.7623202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 50%] 2024-08-06T21:15:51.7624409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 50%] 2024-08-06T21:15:51.7625632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 50%] 2024-08-06T21:15:51.7626827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 50%] 2024-08-06T21:15:51.7628035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 50%] 2024-08-06T21:15:51.7629277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 50%] 2024-08-06T21:15:51.7630530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 50%] 2024-08-06T21:15:51.7631748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 50%] 2024-08-06T21:15:51.7632965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 50%] 2024-08-06T21:15:51.7634171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 50%] 2024-08-06T21:15:51.7635384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 50%] 2024-08-06T21:15:51.7636593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0062s] [ 50%] 2024-08-06T21:15:51.7637799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 50%] 2024-08-06T21:15:51.7639059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 50%] 2024-08-06T21:15:51.7640317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0049s] [ 50%] 2024-08-06T21:15:51.7641521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0070s] [ 50%] 2024-08-06T21:15:51.7642745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 50%] 2024-08-06T21:15:51.7643952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 50%] 2024-08-06T21:15:51.7645171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 50%] 2024-08-06T21:15:51.7646367Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 50%] 2024-08-06T21:15:51.7647615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 51%] 2024-08-06T21:15:51.7648870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 51%] 2024-08-06T21:15:51.7650084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 51%] 2024-08-06T21:15:51.7651283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 51%] 2024-08-06T21:15:51.7652514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 51%] 2024-08-06T21:15:51.7653731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 51%] 2024-08-06T21:15:51.7654940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 51%] 2024-08-06T21:15:51.7656145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0071s] [ 51%] 2024-08-06T21:15:51.7657412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 51%] 2024-08-06T21:15:51.7658692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 51%] 2024-08-06T21:15:51.7659899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 51%] 2024-08-06T21:15:51.7661110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0076s] [ 51%] 2024-08-06T21:15:51.7662342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 51%] 2024-08-06T21:15:51.7663552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 51%] 2024-08-06T21:15:51.7664773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 51%] 2024-08-06T21:15:51.7666012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 51%] 2024-08-06T21:15:51.7667272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 51%] 2024-08-06T21:15:51.7668473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 51%] 2024-08-06T21:15:51.7669681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 51%] 2024-08-06T21:15:51.7670880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 51%] 2024-08-06T21:15:51.7672119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 51%] 2024-08-06T21:15:51.7673318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 51%] 2024-08-06T21:15:51.7674520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 51%] 2024-08-06T21:15:51.7675771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 51%] 2024-08-06T21:15:51.7677029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 51%] 2024-08-06T21:15:51.7678233Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 51%] 2024-08-06T21:15:51.7679434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 51%] 2024-08-06T21:15:51.7680650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 51%] 2024-08-06T21:15:51.7681879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 51%] 2024-08-06T21:15:51.7683091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 51%] 2024-08-06T21:15:51.7684340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 51%] 2024-08-06T21:15:51.7685585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 51%] 2024-08-06T21:15:51.7686796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 51%] 2024-08-06T21:15:51.7687984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 51%] 2024-08-06T21:15:51.7689187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 51%] 2024-08-06T21:15:51.7690455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 51%] 2024-08-06T21:15:51.7691691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 51%] 2024-08-06T21:15:51.7692888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 51%] 2024-08-06T21:15:51.7694147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 51%] 2024-08-06T21:15:51.7695394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 51%] 2024-08-06T21:15:51.7696594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 51%] 2024-08-06T21:15:51.7697797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 51%] 2024-08-06T21:15:51.7699006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0049s] [ 51%] 2024-08-06T21:15:51.7700215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 51%] 2024-08-06T21:15:51.7701443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 51%] 2024-08-06T21:15:51.7702651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 51%] 2024-08-06T21:15:51.7704277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 51%] 2024-08-06T21:15:51.7705632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 51%] 2024-08-06T21:15:51.7706826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 51%] 2024-08-06T21:15:51.7708015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 51%] 2024-08-06T21:15:51.7709221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 51%] 2024-08-06T21:15:51.7710417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 51%] 2024-08-06T21:15:51.7711620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 51%] 2024-08-06T21:15:51.7712811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 51%] 2024-08-06T21:15:51.7714071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 51%] 2024-08-06T21:15:51.7715327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 51%] 2024-08-06T21:15:51.7716543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 51%] 2024-08-06T21:15:51.7717739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 51%] 2024-08-06T21:15:51.7718946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 51%] 2024-08-06T21:15:51.7720149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 51%] 2024-08-06T21:15:51.7721349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 51%] 2024-08-06T21:15:51.7722636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 51%] 2024-08-06T21:15:51.7723920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 51%] 2024-08-06T21:15:51.7725116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 51%] 2024-08-06T21:15:51.7726304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 51%] 2024-08-06T21:15:51.7727502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 51%] 2024-08-06T21:15:51.7728698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 51%] 2024-08-06T21:15:51.7729887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 51%] 2024-08-06T21:15:51.7731096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 51%] 2024-08-06T21:15:51.7732346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 51%] 2024-08-06T21:15:51.7733601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 51%] 2024-08-06T21:15:51.7734800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 51%] 2024-08-06T21:15:51.7736008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 51%] 2024-08-06T21:15:51.7737220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 51%] 2024-08-06T21:15:51.7738421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 51%] 2024-08-06T21:15:51.7739617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 51%] 2024-08-06T21:15:51.7740865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 51%] 2024-08-06T21:15:51.7742123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 52%] 2024-08-06T21:15:51.7743327Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 52%] 2024-08-06T21:15:51.7744524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 52%] 2024-08-06T21:15:51.7745722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 52%] 2024-08-06T21:15:51.7746936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 52%] 2024-08-06T21:15:51.7748122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 52%] 2024-08-06T21:15:51.7749320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 52%] 2024-08-06T21:15:51.7750562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 52%] 2024-08-06T21:15:51.7751812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 52%] 2024-08-06T21:15:51.7753018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 52%] 2024-08-06T21:15:51.7754212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 52%] 2024-08-06T21:15:51.7755418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 52%] 2024-08-06T21:15:51.7756628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 52%] 2024-08-06T21:15:51.7757832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 52%] 2024-08-06T21:15:51.7759065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 52%] 2024-08-06T21:15:51.7760326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 52%] 2024-08-06T21:15:51.7761523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 52%] 2024-08-06T21:15:51.7762719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 52%] 2024-08-06T21:15:51.7763925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 52%] 2024-08-06T21:15:51.7765118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 52%] 2024-08-06T21:15:51.7766333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 52%] 2024-08-06T21:15:51.7767523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 52%] 2024-08-06T21:15:51.7768765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 52%] 2024-08-06T21:15:51.7770007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 52%] 2024-08-06T21:15:51.7771201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 52%] 2024-08-06T21:15:51.7772388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 52%] 2024-08-06T21:15:51.7773594Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0078s] [ 52%] 2024-08-06T21:15:51.7774819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 52%] 2024-08-06T21:15:51.7776042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 52%] 2024-08-06T21:15:51.7777308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 52%] 2024-08-06T21:15:51.7778566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0091s] [ 52%] 2024-08-06T21:15:51.7779787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0091s] [ 52%] 2024-08-06T21:15:51.7780991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 52%] 2024-08-06T21:15:51.7782216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 52%] 2024-08-06T21:15:51.7783419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 52%] 2024-08-06T21:15:51.7784617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 52%] 2024-08-06T21:15:51.7785840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 52%] 2024-08-06T21:15:51.7787095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 52%] 2024-08-06T21:15:51.7788356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 52%] 2024-08-06T21:15:51.7789563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 52%] 2024-08-06T21:15:51.7790772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 52%] 2024-08-06T21:15:51.7791981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 52%] 2024-08-06T21:15:51.7793200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0089s] [ 52%] 2024-08-06T21:15:51.7794405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0090s] [ 52%] 2024-08-06T21:15:51.7795664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 52%] 2024-08-06T21:15:51.7796949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 52%] 2024-08-06T21:15:51.7798159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0100s] [ 52%] 2024-08-06T21:15:51.7799378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0098s] [ 52%] 2024-08-06T21:15:51.7800589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 52%] 2024-08-06T21:15:51.7801813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 52%] 2024-08-06T21:15:51.7803011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 52%] 2024-08-06T21:15:51.7804641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 52%] 2024-08-06T21:15:51.7805964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 52%] 2024-08-06T21:15:51.7807244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 52%] 2024-08-06T21:15:51.7808456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 52%] 2024-08-06T21:15:51.7809663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 52%] 2024-08-06T21:15:51.7810878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 52%] 2024-08-06T21:15:51.7812091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 52%] 2024-08-06T21:15:51.7813299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0104s] [ 52%] 2024-08-06T21:15:51.7814605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 52%] 2024-08-06T21:15:51.7815890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 52%] 2024-08-06T21:15:51.7817103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 52%] 2024-08-06T21:15:51.7818309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0114s] [ 52%] 2024-08-06T21:15:51.7819527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0115s] [ 52%] 2024-08-06T21:15:51.7820738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 52%] 2024-08-06T21:15:51.7821958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 52%] 2024-08-06T21:15:51.7823150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 52%] 2024-08-06T21:15:51.7824402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 52%] 2024-08-06T21:15:51.7825660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 52%] 2024-08-06T21:15:51.7826869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 52%] 2024-08-06T21:15:51.7828065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 52%] 2024-08-06T21:15:51.7829272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 52%] 2024-08-06T21:15:51.7830482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 52%] 2024-08-06T21:15:51.7831685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 52%] 2024-08-06T21:15:51.7832927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0075s] [ 52%] 2024-08-06T21:15:51.7834181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 52%] 2024-08-06T21:15:51.7835394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 52%] 2024-08-06T21:15:51.7836596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 52%] 2024-08-06T21:15:51.7837807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0085s] [ 53%] 2024-08-06T21:15:51.7839023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0085s] [ 53%] 2024-08-06T21:15:51.7840223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 53%] 2024-08-06T21:15:51.7841432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 53%] 2024-08-06T21:15:51.7842667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 53%] 2024-08-06T21:15:51.7843923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 53%] 2024-08-06T21:15:51.7845117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 53%] 2024-08-06T21:15:51.7846328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 53%] 2024-08-06T21:15:51.7847525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 53%] 2024-08-06T21:15:51.7848738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 53%] 2024-08-06T21:15:51.7849931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 53%] 2024-08-06T21:15:51.7851129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 53%] 2024-08-06T21:15:51.7852376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 53%] 2024-08-06T21:15:51.7853636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 53%] 2024-08-06T21:15:51.7854843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 53%] 2024-08-06T21:15:51.7856058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 53%] 2024-08-06T21:15:51.7857271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 53%] 2024-08-06T21:15:51.7858480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 53%] 2024-08-06T21:15:51.7859687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 53%] 2024-08-06T21:15:51.7860937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 53%] 2024-08-06T21:15:51.7862183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 53%] 2024-08-06T21:15:51.7863384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 53%] 2024-08-06T21:15:51.7864573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 53%] 2024-08-06T21:15:51.7865790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 53%] 2024-08-06T21:15:51.7866994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 53%] 2024-08-06T21:15:51.7868221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 53%] 2024-08-06T21:15:51.7869414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 53%] 2024-08-06T21:15:51.7870675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 53%] 2024-08-06T21:15:51.7871924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 53%] 2024-08-06T21:15:51.7873128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 53%] 2024-08-06T21:15:51.7874334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 53%] 2024-08-06T21:15:51.7875542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 53%] 2024-08-06T21:15:51.7876752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 53%] 2024-08-06T21:15:51.7877977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 53%] 2024-08-06T21:15:51.7879187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 53%] 2024-08-06T21:15:51.7880435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 53%] 2024-08-06T21:15:51.7881684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 53%] 2024-08-06T21:15:51.7882878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 53%] 2024-08-06T21:15:51.7884075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 53%] 2024-08-06T21:15:51.7885282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 53%] 2024-08-06T21:15:51.7886479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 53%] 2024-08-06T21:15:51.7887701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 53%] 2024-08-06T21:15:51.7888935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 53%] 2024-08-06T21:15:51.7890642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 53%] 2024-08-06T21:15:51.7891835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 53%] 2024-08-06T21:15:51.7893042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 53%] 2024-08-06T21:15:51.7894247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 53%] 2024-08-06T21:15:51.7895456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 53%] 2024-08-06T21:15:51.7896662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 53%] 2024-08-06T21:15:51.7897885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 53%] 2024-08-06T21:15:51.7899143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 53%] 2024-08-06T21:15:51.7900404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 53%] 2024-08-06T21:15:51.7901604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 53%] 2024-08-06T21:15:51.7902804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 53%] 2024-08-06T21:15:51.7904407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 53%] 2024-08-06T21:15:51.7905639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 53%] 2024-08-06T21:15:51.7906837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 53%] 2024-08-06T21:15:51.7908141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 53%] 2024-08-06T21:15:51.7909407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 53%] 2024-08-06T21:15:51.7910620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 53%] 2024-08-06T21:15:51.7911809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 53%] 2024-08-06T21:15:51.7913015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 53%] 2024-08-06T21:15:51.7914212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 53%] 2024-08-06T21:15:51.7915415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0049s] [ 53%] 2024-08-06T21:15:51.7916607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 53%] 2024-08-06T21:15:51.7917866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 53%] 2024-08-06T21:15:51.7919144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 53%] 2024-08-06T21:15:51.7920349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 53%] 2024-08-06T21:15:51.7921541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 53%] 2024-08-06T21:15:51.7922739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 53%] 2024-08-06T21:15:51.7923931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 53%] 2024-08-06T21:15:51.7925120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 53%] 2024-08-06T21:15:51.7926366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 53%] 2024-08-06T21:15:51.7927615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 53%] 2024-08-06T21:15:51.7928816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 53%] 2024-08-06T21:15:51.7930011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 53%] 2024-08-06T21:15:51.7931213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 53%] 2024-08-06T21:15:51.7932420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0050s] [ 54%] 2024-08-06T21:15:51.7933608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 54%] 2024-08-06T21:15:51.7934808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 54%] 2024-08-06T21:15:51.7936052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 54%] 2024-08-06T21:15:51.7937333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 54%] 2024-08-06T21:15:51.7938546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 54%] 2024-08-06T21:15:51.7939753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 54%] 2024-08-06T21:15:51.7940948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 54%] 2024-08-06T21:15:51.7942143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 54%] 2024-08-06T21:15:51.7943343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 54%] 2024-08-06T21:15:51.7944579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 54%] 2024-08-06T21:15:51.7945847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 54%] 2024-08-06T21:15:51.7947053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 54%] 2024-08-06T21:15:51.7948272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 54%] 2024-08-06T21:15:51.7949490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 54%] 2024-08-06T21:15:51.7950692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 54%] 2024-08-06T21:15:51.7951909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0049s] [ 54%] 2024-08-06T21:15:51.7953103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 54%] 2024-08-06T21:15:51.7954370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 54%] 2024-08-06T21:15:51.7955628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0057s] [ 54%] 2024-08-06T21:15:51.7956853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 54%] 2024-08-06T21:15:51.7958080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 54%] 2024-08-06T21:15:51.7959294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 54%] 2024-08-06T21:15:51.7960507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 54%] 2024-08-06T21:15:51.7961702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 54%] 2024-08-06T21:15:51.7962959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 54%] 2024-08-06T21:15:51.7964210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 54%] 2024-08-06T21:15:51.7965429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 54%] 2024-08-06T21:15:51.7966627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 54%] 2024-08-06T21:15:51.7967865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 54%] 2024-08-06T21:15:51.7969075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 54%] 2024-08-06T21:15:51.7970280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 54%] 2024-08-06T21:15:51.7971493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 54%] 2024-08-06T21:15:51.7972730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 54%] 2024-08-06T21:15:51.7973999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0049s] [ 54%] 2024-08-06T21:15:51.7975205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 54%] 2024-08-06T21:15:51.7976435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 54%] 2024-08-06T21:15:51.7977661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 54%] 2024-08-06T21:15:51.7978888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 54%] 2024-08-06T21:15:51.7980081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 54%] 2024-08-06T21:15:51.7981270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 54%] 2024-08-06T21:15:51.7982517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 54%] 2024-08-06T21:15:51.7983771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 54%] 2024-08-06T21:15:51.7984986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 54%] 2024-08-06T21:15:51.7986195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 54%] 2024-08-06T21:15:51.7987508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 54%] 2024-08-06T21:15:51.7988710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 54%] 2024-08-06T21:15:51.7989970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 54%] 2024-08-06T21:15:51.7991182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 54%] 2024-08-06T21:15:51.7992427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 54%] 2024-08-06T21:15:51.7993692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 54%] 2024-08-06T21:15:51.7994897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 54%] 2024-08-06T21:15:51.7996129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 54%] 2024-08-06T21:15:51.7997342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 54%] 2024-08-06T21:15:51.7998561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 54%] 2024-08-06T21:15:51.7999751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 54%] 2024-08-06T21:15:51.8001001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 54%] 2024-08-06T21:15:51.8002244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 54%] 2024-08-06T21:15:51.8003477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 54%] 2024-08-06T21:15:51.8005054Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 54%] 2024-08-06T21:15:51.8006288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 54%] 2024-08-06T21:15:51.8007498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 54%] 2024-08-06T21:15:51.8008706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 54%] 2024-08-06T21:15:51.8009938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0156s] [ 54%] 2024-08-06T21:15:51.8011257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0166s] [ 54%] 2024-08-06T21:15:51.8012554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0079s] [ 54%] 2024-08-06T21:15:51.8014373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 54%] 2024-08-06T21:15:51.8015655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0233s] [ 54%] 2024-08-06T21:15:51.8016905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0236s] [ 54%] 2024-08-06T21:15:51.8018122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0086s] [ 54%] 2024-08-06T21:15:51.8019369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0086s] [ 54%] 2024-08-06T21:15:51.8020680Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0155s] [ 54%] 2024-08-06T21:15:51.8021994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0160s] [ 54%] 2024-08-06T21:15:51.8023199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0079s] [ 54%] 2024-08-06T21:15:51.8024428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 54%] 2024-08-06T21:15:51.8025651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0226s] [ 54%] 2024-08-06T21:15:51.8026880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0225s] [ 54%] 2024-08-06T21:15:51.8028097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0086s] [ 55%] 2024-08-06T21:15:51.8029331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0085s] [ 55%] 2024-08-06T21:15:51.8030616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0160s] [ 55%] 2024-08-06T21:15:51.8031890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0174s] [ 55%] 2024-08-06T21:15:51.8033124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0080s] [ 55%] 2024-08-06T21:15:51.8034353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 55%] 2024-08-06T21:15:51.8035592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0236s] [ 55%] 2024-08-06T21:15:51.8036812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0238s] [ 55%] 2024-08-06T21:15:51.8038029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0086s] [ 55%] 2024-08-06T21:15:51.8039343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0086s] [ 55%] 2024-08-06T21:15:51.8040682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0161s] [ 55%] 2024-08-06T21:15:51.8041907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0170s] [ 55%] 2024-08-06T21:15:51.8043115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0082s] [ 55%] 2024-08-06T21:15:51.8044345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 55%] 2024-08-06T21:15:51.8045566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0230s] [ 55%] 2024-08-06T21:15:51.8046799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0232s] [ 55%] 2024-08-06T21:15:51.8048018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0087s] [ 55%] 2024-08-06T21:15:51.8049302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0088s] [ 55%] 2024-08-06T21:15:51.8050579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0180s] [ 55%] 2024-08-06T21:15:51.8051810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0197s] [ 55%] 2024-08-06T21:15:51.8053048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0087s] [ 55%] 2024-08-06T21:15:51.8054276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0086s] [ 55%] 2024-08-06T21:15:51.8055508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0252s] [ 55%] 2024-08-06T21:15:51.8056728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0255s] [ 55%] 2024-08-06T21:15:51.8058003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0092s] [ 55%] 2024-08-06T21:15:51.8059307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0095s] [ 55%] 2024-08-06T21:15:51.8060530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0176s] [ 55%] 2024-08-06T21:15:51.8061747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0191s] [ 55%] 2024-08-06T21:15:51.8062962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0085s] [ 55%] 2024-08-06T21:15:51.8064201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0087s] [ 55%] 2024-08-06T21:15:51.8065416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0247s] [ 55%] 2024-08-06T21:15:51.8066656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0248s] [ 55%] 2024-08-06T21:15:51.8067911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0095s] [ 55%] 2024-08-06T21:15:51.8069213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0094s] [ 55%] 2024-08-06T21:15:51.8070429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0151s] [ 55%] 2024-08-06T21:15:51.8071668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0158s] [ 55%] 2024-08-06T21:15:51.8072892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0076s] [ 55%] 2024-08-06T21:15:51.8074108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 55%] 2024-08-06T21:15:51.8075343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0228s] [ 55%] 2024-08-06T21:15:51.8076608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0229s] [ 55%] 2024-08-06T21:15:51.8077909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0085s] [ 55%] 2024-08-06T21:15:51.8079156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 55%] 2024-08-06T21:15:51.8080383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0149s] [ 55%] 2024-08-06T21:15:51.8081606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0154s] [ 55%] 2024-08-06T21:15:51.8082833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0077s] [ 55%] 2024-08-06T21:15:51.8084050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0077s] [ 55%] 2024-08-06T21:15:51.8085254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0221s] [ 55%] 2024-08-06T21:15:51.8086525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0224s] [ 55%] 2024-08-06T21:15:51.8087784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0087s] [ 55%] 2024-08-06T21:15:51.8089036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0086s] [ 55%] 2024-08-06T21:15:51.8090259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0260s] [ 55%] 2024-08-06T21:15:51.8091496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0283s] [ 55%] 2024-08-06T21:15:51.8092707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0097s] [ 55%] 2024-08-06T21:15:51.8093928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0095s] [ 55%] 2024-08-06T21:15:51.8095189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0407s] [ 55%] 2024-08-06T21:15:51.8096469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0416s] [ 55%] 2024-08-06T21:15:51.8097700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0108s] [ 55%] 2024-08-06T21:15:51.8098953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0109s] [ 55%] 2024-08-06T21:15:51.8100171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0232s] [ 55%] 2024-08-06T21:15:51.8101390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0251s] [ 55%] 2024-08-06T21:15:51.8102607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0097s] [ 55%] 2024-08-06T21:15:51.8103863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0097s] [ 55%] 2024-08-06T21:15:51.8105581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0357s] [ 55%] 2024-08-06T21:15:51.8106916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0361s] [ 55%] 2024-08-06T21:15:51.8108135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0110s] [ 55%] 2024-08-06T21:15:51.8109404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0108s] [ 55%] 2024-08-06T21:15:51.8110635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0271s] [ 55%] 2024-08-06T21:15:51.8111874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0302s] [ 55%] 2024-08-06T21:15:51.8113092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0101s] [ 55%] 2024-08-06T21:15:51.8114385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0100s] [ 55%] 2024-08-06T21:15:51.8115674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0415s] [ 55%] 2024-08-06T21:15:51.8116917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0422s] [ 55%] 2024-08-06T21:15:51.8118144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0114s] [ 55%] 2024-08-06T21:15:51.8119398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0112s] [ 55%] 2024-08-06T21:15:51.8120624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0243s] [ 55%] 2024-08-06T21:15:51.8121838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0265s] [ 55%] 2024-08-06T21:15:51.8123109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0102s] [ 55%] 2024-08-06T21:15:51.8124383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0101s] [ 55%] 2024-08-06T21:15:51.8125608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0362s] [ 56%] 2024-08-06T21:15:51.8126827Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0368s] [ 56%] 2024-08-06T21:15:51.8128057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0113s] [ 56%] 2024-08-06T21:15:51.8129311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0113s] [ 56%] 2024-08-06T21:15:51.8130524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0301s] [ 56%] 2024-08-06T21:15:51.8131760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0341s] [ 56%] 2024-08-06T21:15:51.8133028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0107s] [ 56%] 2024-08-06T21:15:51.8134315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0111s] [ 56%] 2024-08-06T21:15:51.8135545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0445s] [ 56%] 2024-08-06T21:15:51.8136793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0448s] [ 56%] 2024-08-06T21:15:51.8138022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0122s] [ 56%] 2024-08-06T21:15:51.8139288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0121s] [ 56%] 2024-08-06T21:15:51.8140495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0269s] [ 56%] 2024-08-06T21:15:51.8141753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0296s] [ 56%] 2024-08-06T21:15:51.8143035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0107s] [ 56%] 2024-08-06T21:15:51.8144251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0110s] [ 56%] 2024-08-06T21:15:51.8145475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0389s] [ 56%] 2024-08-06T21:15:51.8146708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0392s] [ 56%] 2024-08-06T21:15:51.8147937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0119s] [ 56%] 2024-08-06T21:15:51.8149177Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0119s] [ 56%] 2024-08-06T21:15:51.8150398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0249s] [ 56%] 2024-08-06T21:15:51.8151671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0265s] [ 56%] 2024-08-06T21:15:51.8152942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0096s] [ 56%] 2024-08-06T21:15:51.8154173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0093s] [ 56%] 2024-08-06T21:15:51.8155392Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0396s] [ 56%] 2024-08-06T21:15:51.8156648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0405s] [ 56%] 2024-08-06T21:15:51.8157868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0108s] [ 56%] 2024-08-06T21:15:51.8159131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0108s] [ 56%] 2024-08-06T21:15:51.8160383Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0224s] [ 56%] 2024-08-06T21:15:51.8161666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0234s] [ 56%] 2024-08-06T21:15:51.8162874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0095s] [ 56%] 2024-08-06T21:15:51.8164093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0096s] [ 56%] 2024-08-06T21:15:51.8165333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0348s] [ 56%] 2024-08-06T21:15:51.8166558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0352s] [ 56%] 2024-08-06T21:15:51.8167783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0105s] [ 56%] 2024-08-06T21:15:51.8169024Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0104s] [ 56%] 2024-08-06T21:15:51.8170298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0100s] [ 56%] 2024-08-06T21:15:51.8171569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0100s] [ 56%] 2024-08-06T21:15:51.8172795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0067s] [ 56%] 2024-08-06T21:15:51.8174013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 56%] 2024-08-06T21:15:51.8175231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0137s] [ 56%] 2024-08-06T21:15:51.8176461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0136s] [ 56%] 2024-08-06T21:15:51.8177671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0073s] [ 56%] 2024-08-06T21:15:51.8178969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 56%] 2024-08-06T21:15:51.8180232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0100s] [ 56%] 2024-08-06T21:15:51.8181453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0102s] [ 56%] 2024-08-06T21:15:51.8182649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0077s] [ 56%] 2024-08-06T21:15:51.8183874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 56%] 2024-08-06T21:15:51.8185090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0138s] [ 56%] 2024-08-06T21:15:51.8186300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0138s] [ 56%] 2024-08-06T21:15:51.8187515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0075s] [ 56%] 2024-08-06T21:15:51.8188787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 56%] 2024-08-06T21:15:51.8190068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0104s] [ 56%] 2024-08-06T21:15:51.8191278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0104s] [ 56%] 2024-08-06T21:15:51.8192499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0068s] [ 56%] 2024-08-06T21:15:51.8193724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 56%] 2024-08-06T21:15:51.8194947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0141s] [ 56%] 2024-08-06T21:15:51.8196167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0143s] [ 56%] 2024-08-06T21:15:51.8197379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0075s] [ 56%] 2024-08-06T21:15:51.8198677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 56%] 2024-08-06T21:15:51.8199940Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0107s] [ 56%] 2024-08-06T21:15:51.8201160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0106s] [ 56%] 2024-08-06T21:15:51.8202368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0071s] [ 56%] 2024-08-06T21:15:51.8203640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 56%] 2024-08-06T21:15:51.8205257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0143s] [ 56%] 2024-08-06T21:15:51.8206500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0144s] [ 56%] 2024-08-06T21:15:51.8207781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0076s] [ 56%] 2024-08-06T21:15:51.8209089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 56%] 2024-08-06T21:15:51.8210307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0116s] [ 56%] 2024-08-06T21:15:51.8211519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0114s] [ 56%] 2024-08-06T21:15:51.8212747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0074s] [ 56%] 2024-08-06T21:15:51.8213963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 56%] 2024-08-06T21:15:51.8215189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0154s] [ 56%] 2024-08-06T21:15:51.8216401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0152s] [ 56%] 2024-08-06T21:15:51.8217688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0075s] [ 56%] 2024-08-06T21:15:51.8219005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 56%] 2024-08-06T21:15:51.8220207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0118s] [ 56%] 2024-08-06T21:15:51.8221432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0118s] [ 57%] 2024-08-06T21:15:51.8222634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0073s] [ 57%] 2024-08-06T21:15:51.8223853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 57%] 2024-08-06T21:15:51.8225059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0156s] [ 57%] 2024-08-06T21:15:51.8226333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0154s] [ 57%] 2024-08-06T21:15:51.8227595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0078s] [ 57%] 2024-08-06T21:15:51.8228831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 57%] 2024-08-06T21:15:51.8230037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0091s] [ 57%] 2024-08-06T21:15:51.8231253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0090s] [ 57%] 2024-08-06T21:15:51.8232479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 57%] 2024-08-06T21:15:51.8233688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 57%] 2024-08-06T21:15:51.8234914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0134s] [ 57%] 2024-08-06T21:15:51.8236168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0128s] [ 57%] 2024-08-06T21:15:51.8237434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0074s] [ 57%] 2024-08-06T21:15:51.8238658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 57%] 2024-08-06T21:15:51.8239881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0094s] [ 57%] 2024-08-06T21:15:51.8241095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0093s] [ 57%] 2024-08-06T21:15:51.8242294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0065s] [ 57%] 2024-08-06T21:15:51.8243504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 57%] 2024-08-06T21:15:51.8244754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0137s] [ 57%] 2024-08-06T21:15:51.8246026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0132s] [ 57%] 2024-08-06T21:15:51.8247230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0076s] [ 57%] 2024-08-06T21:15:51.8248463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 57%] 2024-08-06T21:15:51.8249687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0482s] [ 57%] 2024-08-06T21:15:51.8250924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0536s] [ 57%] 2024-08-06T21:15:51.8252137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0132s] [ 57%] 2024-08-06T21:15:51.8253359Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0134s] [ 57%] 2024-08-06T21:15:51.8254640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.2477s] [ 57%] 2024-08-06T21:15:51.8255918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0795s] [ 57%] 2024-08-06T21:15:51.8257149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0156s] [ 57%] 2024-08-06T21:15:51.8258387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0161s] [ 57%] 2024-08-06T21:15:51.8259624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0328s] [ 57%] 2024-08-06T21:15:51.8260836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0354s] [ 57%] 2024-08-06T21:15:51.8262061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0130s] [ 57%] 2024-08-06T21:15:51.8263321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0132s] [ 57%] 2024-08-06T21:15:51.8264591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0499s] [ 57%] 2024-08-06T21:15:51.8265825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0512s] [ 57%] 2024-08-06T21:15:51.8267040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0149s] [ 57%] 2024-08-06T21:15:51.8268285Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0151s] [ 57%] 2024-08-06T21:15:51.8269507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0505s] [ 57%] 2024-08-06T21:15:51.8270737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0570s] [ 57%] 2024-08-06T21:15:51.8271959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0146s] [ 57%] 2024-08-06T21:15:51.8273244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0149s] [ 57%] 2024-08-06T21:15:51.8274516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0773s] [ 57%] 2024-08-06T21:15:51.8275747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0801s] [ 57%] 2024-08-06T21:15:51.8276977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0171s] [ 57%] 2024-08-06T21:15:51.8278221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0170s] [ 57%] 2024-08-06T21:15:51.8279444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0345s] [ 57%] 2024-08-06T21:15:51.8280655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0375s] [ 57%] 2024-08-06T21:15:51.8281924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0138s] [ 57%] 2024-08-06T21:15:51.8283196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0138s] [ 57%] 2024-08-06T21:15:51.8284429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0514s] [ 57%] 2024-08-06T21:15:51.8285648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0525s] [ 57%] 2024-08-06T21:15:51.8286881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0159s] [ 57%] 2024-08-06T21:15:51.8288121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0160s] [ 57%] 2024-08-06T21:15:51.8289334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0572s] [ 57%] 2024-08-06T21:15:51.8290619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0643s] [ 57%] 2024-08-06T21:15:51.8291897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0166s] [ 57%] 2024-08-06T21:15:51.8293184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0166s] [ 57%] 2024-08-06T21:15:51.8294407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0826s] [ 57%] 2024-08-06T21:15:51.8295665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0844s] [ 57%] 2024-08-06T21:15:51.8296891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0185s] [ 57%] 2024-08-06T21:15:51.8298153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0185s] [ 57%] 2024-08-06T21:15:51.8299362Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0387s] [ 57%] 2024-08-06T21:15:51.8300624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0428s] [ 57%] 2024-08-06T21:15:51.8301897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0151s] [ 57%] 2024-08-06T21:15:51.8303106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0152s] [ 57%] 2024-08-06T21:15:51.8304787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0551s] [ 57%] 2024-08-06T21:15:51.8306049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0555s] [ 57%] 2024-08-06T21:15:51.8307286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0173s] [ 57%] 2024-08-06T21:15:51.8308522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0167s] [ 57%] 2024-08-06T21:15:51.8309750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0459s] [ 57%] 2024-08-06T21:15:51.8311048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0503s] [ 57%] 2024-08-06T21:15:51.8312323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0130s] [ 57%] 2024-08-06T21:15:51.8313551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0130s] [ 57%] 2024-08-06T21:15:51.8314772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0748s] [ 57%] 2024-08-06T21:15:51.8316017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0776s] [ 57%] 2024-08-06T21:15:51.8317231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0153s] [ 57%] 2024-08-06T21:15:51.8318476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0153s] [ 58%] 2024-08-06T21:15:51.8319752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0312s] [ 58%] 2024-08-06T21:15:51.8321063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0329s] [ 58%] 2024-08-06T21:15:51.8322272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0122s] [ 58%] 2024-08-06T21:15:51.8323489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0120s] [ 58%] 2024-08-06T21:15:51.8324716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0490s] [ 58%] 2024-08-06T21:15:51.8325929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0497s] [ 58%] 2024-08-06T21:15:51.8327148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0143s] [ 58%] 2024-08-06T21:15:51.8328375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0143s] [ 58%] 2024-08-06T21:15:51.8329650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0108s] [ 58%] 2024-08-06T21:15:51.8330919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0111s] [ 58%] 2024-08-06T21:15:51.8332144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 58%] 2024-08-06T21:15:51.8333373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 58%] 2024-08-06T21:15:51.8334598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0148s] [ 58%] 2024-08-06T21:15:51.8335838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0151s] [ 58%] 2024-08-06T21:15:51.8337056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0076s] [ 58%] 2024-08-06T21:15:51.8338352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 58%] 2024-08-06T21:15:51.8339618Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0111s] [ 58%] 2024-08-06T21:15:51.8340846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0114s] [ 58%] 2024-08-06T21:15:51.8342055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0073s] [ 58%] 2024-08-06T21:15:51.8343284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 58%] 2024-08-06T21:15:51.8344503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0153s] [ 58%] 2024-08-06T21:15:51.8345716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0154s] [ 58%] 2024-08-06T21:15:51.8346944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0077s] [ 58%] 2024-08-06T21:15:51.8348221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 58%] 2024-08-06T21:15:51.8349501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0109s] [ 58%] 2024-08-06T21:15:51.8350715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0116s] [ 58%] 2024-08-06T21:15:51.8351944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0073s] [ 58%] 2024-08-06T21:15:51.8353164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 58%] 2024-08-06T21:15:51.8354398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0153s] [ 58%] 2024-08-06T21:15:51.8355616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0152s] [ 58%] 2024-08-06T21:15:51.8356876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0076s] [ 58%] 2024-08-06T21:15:51.8358190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0077s] [ 58%] 2024-08-06T21:15:51.8359403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0113s] [ 58%] 2024-08-06T21:15:51.8360626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0118s] [ 58%] 2024-08-06T21:15:51.8361842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0071s] [ 58%] 2024-08-06T21:15:51.8363070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 58%] 2024-08-06T21:15:51.8364277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0149s] [ 58%] 2024-08-06T21:15:51.8365511Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0152s] [ 58%] 2024-08-06T21:15:51.8366772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0080s] [ 58%] 2024-08-06T21:15:51.8368040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 58%] 2024-08-06T21:15:51.8369275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0118s] [ 58%] 2024-08-06T21:15:51.8370499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0126s] [ 58%] 2024-08-06T21:15:51.8371732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0076s] [ 58%] 2024-08-06T21:15:51.8372945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 58%] 2024-08-06T21:15:51.8374169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0160s] [ 58%] 2024-08-06T21:15:51.8375434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0159s] [ 58%] 2024-08-06T21:15:51.8376714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0081s] [ 58%] 2024-08-06T21:15:51.8377945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 58%] 2024-08-06T21:15:51.8379155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0123s] [ 58%] 2024-08-06T21:15:51.8380388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0128s] [ 58%] 2024-08-06T21:15:51.8381597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0076s] [ 58%] 2024-08-06T21:15:51.8382834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 58%] 2024-08-06T21:15:51.8384055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0162s] [ 58%] 2024-08-06T21:15:51.8385337Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0162s] [ 58%] 2024-08-06T21:15:51.8386595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0082s] [ 58%] 2024-08-06T21:15:51.8387843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0083s] [ 58%] 2024-08-06T21:15:51.8389065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0107s] [ 58%] 2024-08-06T21:15:51.8390287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0107s] [ 58%] 2024-08-06T21:15:51.8391515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 58%] 2024-08-06T21:15:51.8392731Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 58%] 2024-08-06T21:15:51.8394011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0146s] [ 58%] 2024-08-06T21:15:51.8395286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0149s] [ 58%] 2024-08-06T21:15:51.8396516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 58%] 2024-08-06T21:15:51.8397743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 58%] 2024-08-06T21:15:51.8398969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0110s] [ 58%] 2024-08-06T21:15:51.8400187Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0108s] [ 58%] 2024-08-06T21:15:51.8401388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 58%] 2024-08-06T21:15:51.8402620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 58%] 2024-08-06T21:15:51.8404263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0151s] [ 58%] 2024-08-06T21:15:51.8405657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0152s] [ 58%] 2024-08-06T21:15:51.8406868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0078s] [ 58%] 2024-08-06T21:15:51.8408110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0076s] [ 58%] 2024-08-06T21:15:51.8409330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0101s] [ 58%] 2024-08-06T21:15:51.8410560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0100s] [ 58%] 2024-08-06T21:15:51.8411768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0068s] [ 58%] 2024-08-06T21:15:51.8413046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 58%] 2024-08-06T21:15:51.8414350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0139s] [ 59%] 2024-08-06T21:15:51.8415577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0141s] [ 59%] 2024-08-06T21:15:51.8416806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0074s] [ 59%] 2024-08-06T21:15:51.8418040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 59%] 2024-08-06T21:15:51.8419268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0105s] [ 59%] 2024-08-06T21:15:51.8420476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 59%] 2024-08-06T21:15:51.8421700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0069s] [ 59%] 2024-08-06T21:15:51.8422971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 59%] 2024-08-06T21:15:51.8424241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0143s] [ 59%] 2024-08-06T21:15:51.8425472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0143s] [ 59%] 2024-08-06T21:15:51.8426686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0075s] [ 59%] 2024-08-06T21:15:51.8427924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0076s] [ 59%] 2024-08-06T21:15:51.8429135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0094s] [ 59%] 2024-08-06T21:15:51.8430368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0106s] [ 59%] 2024-08-06T21:15:51.8431631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 59%] 2024-08-06T21:15:51.8432925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 59%] 2024-08-06T21:15:51.8434140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0145s] [ 59%] 2024-08-06T21:15:51.8435353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0145s] [ 59%] 2024-08-06T21:15:51.8436589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 59%] 2024-08-06T21:15:51.8437819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0076s] [ 59%] 2024-08-06T21:15:51.8439040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0098s] [ 59%] 2024-08-06T21:15:51.8440251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0108s] [ 59%] 2024-08-06T21:15:51.8441512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0071s] [ 59%] 2024-08-06T21:15:51.8442770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 59%] 2024-08-06T21:15:51.8443993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0149s] [ 59%] 2024-08-06T21:15:51.8445210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0150s] [ 59%] 2024-08-06T21:15:51.8446424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0072s] [ 59%] 2024-08-06T21:15:51.8447670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0076s] [ 59%] 2024-08-06T21:15:51.8448880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0116s] [ 59%] 2024-08-06T21:15:51.8450159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0117s] [ 59%] 2024-08-06T21:15:51.8451423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0073s] [ 59%] 2024-08-06T21:15:51.8452661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 59%] 2024-08-06T21:15:51.8453874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0156s] [ 59%] 2024-08-06T21:15:51.8455120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0158s] [ 59%] 2024-08-06T21:15:51.8456340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0078s] [ 59%] 2024-08-06T21:15:51.8457572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 59%] 2024-08-06T21:15:51.8458796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0121s] [ 59%] 2024-08-06T21:15:51.8460055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0121s] [ 59%] 2024-08-06T21:15:51.8461318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0075s] [ 59%] 2024-08-06T21:15:51.8462524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 59%] 2024-08-06T21:15:51.8463752Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0158s] [ 59%] 2024-08-06T21:15:51.8464981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0158s] [ 59%] 2024-08-06T21:15:51.8466216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0087s] [ 59%] 2024-08-06T21:15:51.8467445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 59%] 2024-08-06T21:15:51.8468701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0099s] [ 59%] 2024-08-06T21:15:51.8469988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0096s] [ 59%] 2024-08-06T21:15:51.8471192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0066s] [ 59%] 2024-08-06T21:15:51.8472415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 59%] 2024-08-06T21:15:51.8473637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0139s] [ 59%] 2024-08-06T21:15:51.8474874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0139s] [ 59%] 2024-08-06T21:15:51.8476083Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0073s] [ 59%] 2024-08-06T21:15:51.8477330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 59%] 2024-08-06T21:15:51.8478577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0098s] [ 59%] 2024-08-06T21:15:51.8479839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0100s] [ 59%] 2024-08-06T21:15:51.8481060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0069s] [ 59%] 2024-08-06T21:15:51.8482273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 59%] 2024-08-06T21:15:51.8483500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0140s] [ 59%] 2024-08-06T21:15:51.8484715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0141s] [ 59%] 2024-08-06T21:15:51.8485931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0075s] [ 59%] 2024-08-06T21:15:51.8487202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 59%] 2024-08-06T21:15:51.8488487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 59%] 2024-08-06T21:15:51.8489710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 59%] 2024-08-06T21:15:51.8490920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 59%] 2024-08-06T21:15:51.8492164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 59%] 2024-08-06T21:15:51.8493387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 59%] 2024-08-06T21:15:51.8494629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 59%] 2024-08-06T21:15:51.8495843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 59%] 2024-08-06T21:15:51.8497150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 59%] 2024-08-06T21:15:51.8498531Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 59%] 2024-08-06T21:15:51.8499765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 59%] 2024-08-06T21:15:51.8500970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 59%] 2024-08-06T21:15:51.8502193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 59%] 2024-08-06T21:15:51.8503425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 59%] 2024-08-06T21:15:51.8505016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 59%] 2024-08-06T21:15:51.8506364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 59%] 2024-08-06T21:15:51.8507677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 59%] 2024-08-06T21:15:51.8508909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 59%] 2024-08-06T21:15:51.8510128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 60%] 2024-08-06T21:15:51.8511360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 60%] 2024-08-06T21:15:51.8512584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 60%] 2024-08-06T21:15:51.8513802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0075s] [ 60%] 2024-08-06T21:15:51.8515046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 60%] 2024-08-06T21:15:51.8516319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 60%] 2024-08-06T21:15:51.8517647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 60%] 2024-08-06T21:15:51.8518856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 60%] 2024-08-06T21:15:51.8520085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 60%] 2024-08-06T21:15:51.8521298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 60%] 2024-08-06T21:15:51.8522524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 60%] 2024-08-06T21:15:51.8523736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 60%] 2024-08-06T21:15:51.8525011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 60%] 2024-08-06T21:15:51.8526278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 60%] 2024-08-06T21:15:51.8527516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 60%] 2024-08-06T21:15:51.8528739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 60%] 2024-08-06T21:15:51.8529960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 60%] 2024-08-06T21:15:51.8531188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 60%] 2024-08-06T21:15:51.8532407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 60%] 2024-08-06T21:15:51.8533636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0076s] [ 60%] 2024-08-06T21:15:51.8534900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 60%] 2024-08-06T21:15:51.8536193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 60%] 2024-08-06T21:15:51.8537435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 60%] 2024-08-06T21:15:51.8538647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 60%] 2024-08-06T21:15:51.8539882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 60%] 2024-08-06T21:15:51.8541092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 60%] 2024-08-06T21:15:51.8542319Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 60%] 2024-08-06T21:15:51.8543572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0069s] [ 60%] 2024-08-06T21:15:51.8544863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 60%] 2024-08-06T21:15:51.8546069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 60%] 2024-08-06T21:15:51.8547324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 60%] 2024-08-06T21:15:51.8548542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 60%] 2024-08-06T21:15:51.8549770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 60%] 2024-08-06T21:15:51.8550991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 60%] 2024-08-06T21:15:51.8552209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 60%] 2024-08-06T21:15:51.8553490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 60%] 2024-08-06T21:15:51.8554760Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0070s] [ 60%] 2024-08-06T21:15:51.8555986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 60%] 2024-08-06T21:15:51.8557232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 60%] 2024-08-06T21:15:51.8558457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 60%] 2024-08-06T21:15:51.8559673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 60%] 2024-08-06T21:15:51.8560876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 60%] 2024-08-06T21:15:51.8562142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 60%] 2024-08-06T21:15:51.8563408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 60%] 2024-08-06T21:15:51.8564643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 60%] 2024-08-06T21:15:51.8565851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 60%] 2024-08-06T21:15:51.8567109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 60%] 2024-08-06T21:15:51.8568328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0082s] [ 60%] 2024-08-06T21:15:51.8569556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 60%] 2024-08-06T21:15:51.8570768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 60%] 2024-08-06T21:15:51.8572041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 60%] 2024-08-06T21:15:51.8573330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0102s] [ 60%] 2024-08-06T21:15:51.8574555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 60%] 2024-08-06T21:15:51.8575794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 60%] 2024-08-06T21:15:51.8577048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 60%] 2024-08-06T21:15:51.8578278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 60%] 2024-08-06T21:15:51.8579491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 60%] 2024-08-06T21:15:51.8580754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 60%] 2024-08-06T21:15:51.8582025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 60%] 2024-08-06T21:15:51.8583243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0071s] [ 60%] 2024-08-06T21:15:51.8584472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 60%] 2024-08-06T21:15:51.8585683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 60%] 2024-08-06T21:15:51.8586943Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 60%] 2024-08-06T21:15:51.8588158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0083s] [ 60%] 2024-08-06T21:15:51.8589390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0087s] [ 60%] 2024-08-06T21:15:51.8590700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 60%] 2024-08-06T21:15:51.8591999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 60%] 2024-08-06T21:15:51.8593216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0104s] [ 60%] 2024-08-06T21:15:51.8594442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0106s] [ 60%] 2024-08-06T21:15:51.8595684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 60%] 2024-08-06T21:15:51.8596933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 60%] 2024-08-06T21:15:51.8598157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 60%] 2024-08-06T21:15:51.8599415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 60%] 2024-08-06T21:15:51.8600685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 60%] 2024-08-06T21:15:51.8601905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 60%] 2024-08-06T21:15:51.8603129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0070s] [ 60%] 2024-08-06T21:15:51.8604721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 60%] 2024-08-06T21:15:51.8606070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 60%] 2024-08-06T21:15:51.8607329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 61%] 2024-08-06T21:15:51.8608545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0092s] [ 61%] 2024-08-06T21:15:51.8609855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0095s] [ 61%] 2024-08-06T21:15:51.8611139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 61%] 2024-08-06T21:15:51.8612381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 61%] 2024-08-06T21:15:51.8613604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0109s] [ 61%] 2024-08-06T21:15:51.8614854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0110s] [ 61%] 2024-08-06T21:15:51.8616075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 61%] 2024-08-06T21:15:51.8617321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 61%] 2024-08-06T21:15:51.8618612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0067s] [ 61%] 2024-08-06T21:15:51.8619941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 61%] 2024-08-06T21:15:51.8621170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 61%] 2024-08-06T21:15:51.8622384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 61%] 2024-08-06T21:15:51.8623615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 61%] 2024-08-06T21:15:51.8624838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 61%] 2024-08-06T21:15:51.8626059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 61%] 2024-08-06T21:15:51.8627304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 61%] 2024-08-06T21:15:51.8628560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0075s] [ 61%] 2024-08-06T21:15:51.8629848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 61%] 2024-08-06T21:15:51.8631058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 61%] 2024-08-06T21:15:51.8632294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 61%] 2024-08-06T21:15:51.8633516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0101s] [ 61%] 2024-08-06T21:15:51.8634758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 61%] 2024-08-06T21:15:51.8635968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 61%] 2024-08-06T21:15:51.8637267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 61%] 2024-08-06T21:15:51.8638537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 61%] 2024-08-06T21:15:51.8639751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 61%] 2024-08-06T21:15:51.8640971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 61%] 2024-08-06T21:15:51.8642185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 61%] 2024-08-06T21:15:51.8643417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0071s] [ 61%] 2024-08-06T21:15:51.8644632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 61%] 2024-08-06T21:15:51.8645855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 61%] 2024-08-06T21:15:51.8647134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 61%] 2024-08-06T21:15:51.8648419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 61%] 2024-08-06T21:15:51.8649629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 61%] 2024-08-06T21:15:51.8650832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 61%] 2024-08-06T21:15:51.8652066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 61%] 2024-08-06T21:15:51.8653275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 61%] 2024-08-06T21:15:51.8654508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 61%] 2024-08-06T21:15:51.8655755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 61%] 2024-08-06T21:15:51.8657059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 61%] 2024-08-06T21:15:51.8658260Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 61%] 2024-08-06T21:15:51.8659474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 61%] 2024-08-06T21:15:51.8660676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 61%] 2024-08-06T21:15:51.8661887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 61%] 2024-08-06T21:15:51.8663108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 61%] 2024-08-06T21:15:51.8664310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 61%] 2024-08-06T21:15:51.8665572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 61%] 2024-08-06T21:15:51.8666840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 61%] 2024-08-06T21:15:51.8668059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 61%] 2024-08-06T21:15:51.8669268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0050s] [ 61%] 2024-08-06T21:15:51.8670493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 61%] 2024-08-06T21:15:51.8671713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0049s] [ 61%] 2024-08-06T21:15:51.8672921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 61%] 2024-08-06T21:15:51.8674198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 61%] 2024-08-06T21:15:51.8675464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 61%] 2024-08-06T21:15:51.8676724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 61%] 2024-08-06T21:15:51.8677925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 61%] 2024-08-06T21:15:51.8679150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 61%] 2024-08-06T21:15:51.8680351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 61%] 2024-08-06T21:15:51.8681571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 61%] 2024-08-06T21:15:51.8682772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 61%] 2024-08-06T21:15:51.8684030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 61%] 2024-08-06T21:15:51.8685304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 61%] 2024-08-06T21:15:51.8686512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 61%] 2024-08-06T21:15:51.8687735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 61%] 2024-08-06T21:15:51.8688952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 61%] 2024-08-06T21:15:51.8690175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 61%] 2024-08-06T21:15:51.8691384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 61%] 2024-08-06T21:15:51.8692602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 61%] 2024-08-06T21:15:51.8693861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 61%] 2024-08-06T21:15:51.8695170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 61%] 2024-08-06T21:15:51.8696408Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 61%] 2024-08-06T21:15:51.8697627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 61%] 2024-08-06T21:15:51.8698854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 61%] 2024-08-06T21:15:51.8700052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 61%] 2024-08-06T21:15:51.8701268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 61%] 2024-08-06T21:15:51.8702510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 62%] 2024-08-06T21:15:51.8703821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 62%] 2024-08-06T21:15:51.8705350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 62%] 2024-08-06T21:15:51.8706574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 62%] 2024-08-06T21:15:51.8707816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 62%] 2024-08-06T21:15:51.8709026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 62%] 2024-08-06T21:15:51.8710236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 62%] 2024-08-06T21:15:51.8711437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 62%] 2024-08-06T21:15:51.8712744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 62%] 2024-08-06T21:15:51.8714028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 62%] 2024-08-06T21:15:51.8715245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 62%] 2024-08-06T21:15:51.8716458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 62%] 2024-08-06T21:15:51.8717676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 62%] 2024-08-06T21:15:51.8718894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 62%] 2024-08-06T21:15:51.8720084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 62%] 2024-08-06T21:15:51.8721300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 62%] 2024-08-06T21:15:51.8722554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 62%] 2024-08-06T21:15:51.8723838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 62%] 2024-08-06T21:15:51.8725034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 62%] 2024-08-06T21:15:51.8726257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 62%] 2024-08-06T21:15:51.8727503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0124s] [ 62%] 2024-08-06T21:15:51.8728725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0133s] [ 62%] 2024-08-06T21:15:51.8729956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0068s] [ 62%] 2024-08-06T21:15:51.8731211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 62%] 2024-08-06T21:15:51.8732502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0165s] [ 62%] 2024-08-06T21:15:51.8733726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0169s] [ 62%] 2024-08-06T21:15:51.8734956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 62%] 2024-08-06T21:15:51.8736184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 62%] 2024-08-06T21:15:51.8737434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0076s] [ 62%] 2024-08-06T21:15:51.8738648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 62%] 2024-08-06T21:15:51.8739856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0066s] [ 62%] 2024-08-06T21:15:51.8741133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 62%] 2024-08-06T21:15:51.8742395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0082s] [ 62%] 2024-08-06T21:15:51.8743626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0083s] [ 62%] 2024-08-06T21:15:51.8744842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 62%] 2024-08-06T21:15:51.8746079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 62%] 2024-08-06T21:15:51.8747317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0130s] [ 62%] 2024-08-06T21:15:51.8748556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0138s] [ 62%] 2024-08-06T21:15:51.8749811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0077s] [ 62%] 2024-08-06T21:15:51.8751080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 62%] 2024-08-06T21:15:51.8752316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0168s] [ 62%] 2024-08-06T21:15:51.8753536Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0168s] [ 62%] 2024-08-06T21:15:51.8754774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0076s] [ 62%] 2024-08-06T21:15:51.8756005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0076s] [ 62%] 2024-08-06T21:15:51.8757266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0071s] [ 62%] 2024-08-06T21:15:51.8758479Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 62%] 2024-08-06T21:15:51.8759745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0067s] [ 62%] 2024-08-06T21:15:51.8761006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 62%] 2024-08-06T21:15:51.8762220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0081s] [ 62%] 2024-08-06T21:15:51.8763460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0086s] [ 62%] 2024-08-06T21:15:51.8764677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 62%] 2024-08-06T21:15:51.8765911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 62%] 2024-08-06T21:15:51.8767143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0142s] [ 62%] 2024-08-06T21:15:51.8768425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0151s] [ 62%] 2024-08-06T21:15:51.8769688Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 62%] 2024-08-06T21:15:51.8770918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 62%] 2024-08-06T21:15:51.8772132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0179s] [ 62%] 2024-08-06T21:15:51.8773364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0181s] [ 62%] 2024-08-06T21:15:51.8774603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 62%] 2024-08-06T21:15:51.8775821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 62%] 2024-08-06T21:15:51.8777068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0082s] [ 62%] 2024-08-06T21:15:51.8778323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0083s] [ 62%] 2024-08-06T21:15:51.8779608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 62%] 2024-08-06T21:15:51.8780819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 62%] 2024-08-06T21:15:51.8782051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0087s] [ 62%] 2024-08-06T21:15:51.8783277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0088s] [ 62%] 2024-08-06T21:15:51.8784481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0075s] [ 62%] 2024-08-06T21:15:51.8785710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0077s] [ 62%] 2024-08-06T21:15:51.8786984Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0122s] [ 62%] 2024-08-06T21:15:51.8788275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0125s] [ 62%] 2024-08-06T21:15:51.8789482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0067s] [ 62%] 2024-08-06T21:15:51.8790715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 62%] 2024-08-06T21:15:51.8791931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0163s] [ 62%] 2024-08-06T21:15:51.8793176Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0168s] [ 62%] 2024-08-06T21:15:51.8794386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0074s] [ 62%] 2024-08-06T21:15:51.8795609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 62%] 2024-08-06T21:15:51.8796903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0073s] [ 62%] 2024-08-06T21:15:51.8798166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 62%] 2024-08-06T21:15:51.8799380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0065s] [ 63%] 2024-08-06T21:15:51.8800596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 63%] 2024-08-06T21:15:51.8801823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0084s] [ 63%] 2024-08-06T21:15:51.8803037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0085s] [ 63%] 2024-08-06T21:15:51.8804607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 63%] 2024-08-06T21:15:51.8805926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 63%] 2024-08-06T21:15:51.8807232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 63%] 2024-08-06T21:15:51.8808466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 63%] 2024-08-06T21:15:51.8809670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 63%] 2024-08-06T21:15:51.8810904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 63%] 2024-08-06T21:15:51.8812125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 63%] 2024-08-06T21:15:51.8813361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 63%] 2024-08-06T21:15:51.8814578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 63%] 2024-08-06T21:15:51.8815874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 63%] 2024-08-06T21:15:51.8817161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 63%] 2024-08-06T21:15:51.8818373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 63%] 2024-08-06T21:15:51.8819596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 63%] 2024-08-06T21:15:51.8820810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 63%] 2024-08-06T21:15:51.8822043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 63%] 2024-08-06T21:15:51.8823257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 63%] 2024-08-06T21:15:51.8824522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 63%] 2024-08-06T21:15:51.8825789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 63%] 2024-08-06T21:15:51.8827034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 63%] 2024-08-06T21:15:51.8828252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 63%] 2024-08-06T21:15:51.8829468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 63%] 2024-08-06T21:15:51.8830711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 63%] 2024-08-06T21:15:51.8831930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 63%] 2024-08-06T21:15:51.8833169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 63%] 2024-08-06T21:15:51.8834427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 63%] 2024-08-06T21:15:51.8835719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 63%] 2024-08-06T21:15:51.8836942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 63%] 2024-08-06T21:15:51.8838171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 63%] 2024-08-06T21:15:51.8839381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 63%] 2024-08-06T21:15:51.8840593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 63%] 2024-08-06T21:15:51.8841815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 63%] 2024-08-06T21:15:51.8843072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 63%] 2024-08-06T21:15:51.8844347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 63%] 2024-08-06T21:15:51.8845564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 63%] 2024-08-06T21:15:51.8846805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 63%] 2024-08-06T21:15:51.8848025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 63%] 2024-08-06T21:15:51.8849257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 63%] 2024-08-06T21:15:51.8850470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 63%] 2024-08-06T21:15:51.8851684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 63%] 2024-08-06T21:15:51.8853000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 63%] 2024-08-06T21:15:51.8854280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 63%] 2024-08-06T21:15:51.8855513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 63%] 2024-08-06T21:15:51.8856727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 63%] 2024-08-06T21:15:51.8857960Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 63%] 2024-08-06T21:15:51.8859159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 63%] 2024-08-06T21:15:51.8860375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 63%] 2024-08-06T21:15:51.8861625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0068s] [ 63%] 2024-08-06T21:15:51.8862892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 63%] 2024-08-06T21:15:51.8864112Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 63%] 2024-08-06T21:15:51.8865321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 63%] 2024-08-06T21:15:51.8866558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 63%] 2024-08-06T21:15:51.8867774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 63%] 2024-08-06T21:15:51.8868995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 63%] 2024-08-06T21:15:51.8870198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 63%] 2024-08-06T21:15:51.8871465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 63%] 2024-08-06T21:15:51.8872732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 63%] 2024-08-06T21:15:51.8873937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 63%] 2024-08-06T21:15:51.8875170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 63%] 2024-08-06T21:15:51.8876389Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 63%] 2024-08-06T21:15:51.8877611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 63%] 2024-08-06T21:15:51.8878805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 63%] 2024-08-06T21:15:51.8880065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 63%] 2024-08-06T21:15:51.8881320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 63%] 2024-08-06T21:15:51.8882545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 63%] 2024-08-06T21:15:51.8883739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 63%] 2024-08-06T21:15:51.8884950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 63%] 2024-08-06T21:15:51.8886182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0049s] [ 63%] 2024-08-06T21:15:51.8887406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0050s] [ 63%] 2024-08-06T21:15:51.8888624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 63%] 2024-08-06T21:15:51.8889926Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 63%] 2024-08-06T21:15:51.8891213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 63%] 2024-08-06T21:15:51.8892426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 63%] 2024-08-06T21:15:51.8893652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 63%] 2024-08-06T21:15:51.8894877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 64%] 2024-08-06T21:15:51.8896077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 64%] 2024-08-06T21:15:51.8897309Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 64%] 2024-08-06T21:15:51.8898546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 64%] 2024-08-06T21:15:51.8899823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 64%] 2024-08-06T21:15:51.8901027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 64%] 2024-08-06T21:15:51.8902252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 64%] 2024-08-06T21:15:51.8903457Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 64%] 2024-08-06T21:15:51.8905058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 64%] 2024-08-06T21:15:51.8906286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 64%] 2024-08-06T21:15:51.8907528Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 64%] 2024-08-06T21:15:51.8908828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 64%] 2024-08-06T21:15:51.8910109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 64%] 2024-08-06T21:15:51.8911334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 64%] 2024-08-06T21:15:51.8912544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 64%] 2024-08-06T21:15:51.8913772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 64%] 2024-08-06T21:15:51.8914990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 64%] 2024-08-06T21:15:51.8916204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 64%] 2024-08-06T21:15:51.8917495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 64%] 2024-08-06T21:15:51.8918787Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 64%] 2024-08-06T21:15:51.8920011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 64%] 2024-08-06T21:15:51.8921213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 64%] 2024-08-06T21:15:51.8922435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 64%] 2024-08-06T21:15:51.8923639Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 64%] 2024-08-06T21:15:51.8924862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 64%] 2024-08-06T21:15:51.8926064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 64%] 2024-08-06T21:15:51.8927353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 64%] 2024-08-06T21:15:51.8928612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 64%] 2024-08-06T21:15:51.8929822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 64%] 2024-08-06T21:15:51.8931043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 64%] 2024-08-06T21:15:51.8932262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 64%] 2024-08-06T21:15:51.8933495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 64%] 2024-08-06T21:15:51.8934711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 64%] 2024-08-06T21:15:51.8935974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 64%] 2024-08-06T21:15:51.8937255Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 64%] 2024-08-06T21:15:51.8938475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 64%] 2024-08-06T21:15:51.8939673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 64%] 2024-08-06T21:15:51.8940877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 64%] 2024-08-06T21:15:51.8942110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 64%] 2024-08-06T21:15:51.8943313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 64%] 2024-08-06T21:15:51.8944537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 64%] 2024-08-06T21:15:51.8945782Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 64%] 2024-08-06T21:15:51.8947082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 64%] 2024-08-06T21:15:51.8948282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 64%] 2024-08-06T21:15:51.8949499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 64%] 2024-08-06T21:15:51.8950709Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 64%] 2024-08-06T21:15:51.8951923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 64%] 2024-08-06T21:15:51.8953146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 64%] 2024-08-06T21:15:51.8954355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 64%] 2024-08-06T21:15:51.8955610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 64%] 2024-08-06T21:15:51.8956880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 64%] 2024-08-06T21:15:51.8958093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 64%] 2024-08-06T21:15:51.8959293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 64%] 2024-08-06T21:15:51.8960515Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 64%] 2024-08-06T21:15:51.8961720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 64%] 2024-08-06T21:15:51.8962915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 64%] 2024-08-06T21:15:51.8964138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 64%] 2024-08-06T21:15:51.8965399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 64%] 2024-08-06T21:15:51.8966708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 64%] 2024-08-06T21:15:51.8967921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 64%] 2024-08-06T21:15:51.8969158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 64%] 2024-08-06T21:15:51.8970375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 64%] 2024-08-06T21:15:51.8971614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 64%] 2024-08-06T21:15:51.8972826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0055s] [ 64%] 2024-08-06T21:15:51.8974088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 64%] 2024-08-06T21:15:51.8975358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 64%] 2024-08-06T21:15:51.8976585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 64%] 2024-08-06T21:15:51.8977808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 64%] 2024-08-06T21:15:51.8979022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 64%] 2024-08-06T21:15:51.8980253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 64%] 2024-08-06T21:15:51.8981465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 64%] 2024-08-06T21:15:51.8982687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 64%] 2024-08-06T21:15:51.8983944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 64%] 2024-08-06T21:15:51.8985204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 64%] 2024-08-06T21:15:51.8986446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 64%] 2024-08-06T21:15:51.8987662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 64%] 2024-08-06T21:15:51.8988907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 64%] 2024-08-06T21:15:51.8990118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0057s] [ 65%] 2024-08-06T21:15:51.8991353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 65%] 2024-08-06T21:15:51.8992655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 65%] 2024-08-06T21:15:51.8993951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 65%] 2024-08-06T21:15:51.8995152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 65%] 2024-08-06T21:15:51.8996364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 65%] 2024-08-06T21:15:51.8997587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 65%] 2024-08-06T21:15:51.8998803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 65%] 2024-08-06T21:15:51.9000026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 65%] 2024-08-06T21:15:51.9001240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 65%] 2024-08-06T21:15:51.9002513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 65%] 2024-08-06T21:15:51.9003812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 65%] 2024-08-06T21:15:51.9005395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 65%] 2024-08-06T21:15:51.9006646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 65%] 2024-08-06T21:15:51.9007862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 65%] 2024-08-06T21:15:51.9009099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 65%] 2024-08-06T21:15:51.9010312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 65%] 2024-08-06T21:15:51.9011625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 65%] 2024-08-06T21:15:51.9012910Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 65%] 2024-08-06T21:15:51.9014148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 65%] 2024-08-06T21:15:51.9015355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 65%] 2024-08-06T21:15:51.9016603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 65%] 2024-08-06T21:15:51.9017814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 65%] 2024-08-06T21:15:51.9019023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 65%] 2024-08-06T21:15:51.9020252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 65%] 2024-08-06T21:15:51.9021523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 65%] 2024-08-06T21:15:51.9022807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 65%] 2024-08-06T21:15:51.9024022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 65%] 2024-08-06T21:15:51.9025247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 65%] 2024-08-06T21:15:51.9026464Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 65%] 2024-08-06T21:15:51.9027684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 65%] 2024-08-06T21:15:51.9028896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 65%] 2024-08-06T21:15:51.9030147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0057s] [ 65%] 2024-08-06T21:15:51.9031439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 65%] 2024-08-06T21:15:51.9032648Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 65%] 2024-08-06T21:15:51.9033879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 65%] 2024-08-06T21:15:51.9035088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 65%] 2024-08-06T21:15:51.9036317Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 65%] 2024-08-06T21:15:51.9037524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 65%] 2024-08-06T21:15:51.9038745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 65%] 2024-08-06T21:15:51.9039994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 65%] 2024-08-06T21:15:51.9041259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 65%] 2024-08-06T21:15:51.9042481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 65%] 2024-08-06T21:15:51.9043694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 65%] 2024-08-06T21:15:51.9044928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 65%] 2024-08-06T21:15:51.9046150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 65%] 2024-08-06T21:15:51.9047388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 65%] 2024-08-06T21:15:51.9048645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0049s] [ 65%] 2024-08-06T21:15:51.9049925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 65%] 2024-08-06T21:15:51.9051147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 65%] 2024-08-06T21:15:51.9052360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 65%] 2024-08-06T21:15:51.9053600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 65%] 2024-08-06T21:15:51.9054813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 65%] 2024-08-06T21:15:51.9056035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 65%] 2024-08-06T21:15:51.9057257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 65%] 2024-08-06T21:15:51.9058527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 65%] 2024-08-06T21:15:51.9059802Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 65%] 2024-08-06T21:15:51.9061034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 65%] 2024-08-06T21:15:51.9062246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 65%] 2024-08-06T21:15:51.9063462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 65%] 2024-08-06T21:15:51.9064695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 65%] 2024-08-06T21:15:51.9065913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 65%] 2024-08-06T21:15:51.9067200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 65%] 2024-08-06T21:15:51.9068466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 65%] 2024-08-06T21:15:51.9069696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0072s] [ 65%] 2024-08-06T21:15:51.9070918Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 65%] 2024-08-06T21:15:51.9072151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 65%] 2024-08-06T21:15:51.9073377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 65%] 2024-08-06T21:15:51.9074580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 65%] 2024-08-06T21:15:51.9075803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 65%] 2024-08-06T21:15:51.9077067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 65%] 2024-08-06T21:15:51.9078352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 65%] 2024-08-06T21:15:51.9079560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 65%] 2024-08-06T21:15:51.9080991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 65%] 2024-08-06T21:15:51.9082217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 65%] 2024-08-06T21:15:51.9083451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 65%] 2024-08-06T21:15:51.9084660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0072s] [ 65%] 2024-08-06T21:15:51.9085917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 65%] 2024-08-06T21:15:51.9087217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 66%] 2024-08-06T21:15:51.9088436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 66%] 2024-08-06T21:15:51.9089666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 66%] 2024-08-06T21:15:51.9090890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0078s] [ 66%] 2024-08-06T21:15:51.9092129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 66%] 2024-08-06T21:15:51.9093344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 66%] 2024-08-06T21:15:51.9094568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 66%] 2024-08-06T21:15:51.9095829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 66%] 2024-08-06T21:15:51.9097105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 66%] 2024-08-06T21:15:51.9098330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 66%] 2024-08-06T21:15:51.9099535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 66%] 2024-08-06T21:15:51.9100769Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 66%] 2024-08-06T21:15:51.9101977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 66%] 2024-08-06T21:15:51.9103201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 66%] 2024-08-06T21:15:51.9104843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 66%] 2024-08-06T21:15:51.9106178Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 66%] 2024-08-06T21:15:51.9107407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 66%] 2024-08-06T21:15:51.9108616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 66%] 2024-08-06T21:15:51.9109843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 66%] 2024-08-06T21:15:51.9111064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 66%] 2024-08-06T21:15:51.9112290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 66%] 2024-08-06T21:15:51.9113506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 66%] 2024-08-06T21:15:51.9114784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 66%] 2024-08-06T21:15:51.9116057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 66%] 2024-08-06T21:15:51.9117299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 66%] 2024-08-06T21:15:51.9118506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 66%] 2024-08-06T21:15:51.9119713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 66%] 2024-08-06T21:15:51.9120942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 66%] 2024-08-06T21:15:51.9122143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 66%] 2024-08-06T21:15:51.9123416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 66%] 2024-08-06T21:15:51.9124667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 66%] 2024-08-06T21:15:51.9125893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 66%] 2024-08-06T21:15:51.9127115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 66%] 2024-08-06T21:15:51.9128343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 66%] 2024-08-06T21:15:51.9129552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 66%] 2024-08-06T21:15:51.9130757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 66%] 2024-08-06T21:15:51.9131973Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 66%] 2024-08-06T21:15:51.9133237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 66%] 2024-08-06T21:15:51.9134499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 66%] 2024-08-06T21:15:51.9135697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 66%] 2024-08-06T21:15:51.9136924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 66%] 2024-08-06T21:15:51.9138127Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 66%] 2024-08-06T21:15:51.9139348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 66%] 2024-08-06T21:15:51.9140545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 66%] 2024-08-06T21:15:51.9141738Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 66%] 2024-08-06T21:15:51.9143004Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 66%] 2024-08-06T21:15:51.9144258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 66%] 2024-08-06T21:15:51.9145482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 66%] 2024-08-06T21:15:51.9146695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 66%] 2024-08-06T21:15:51.9147924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 66%] 2024-08-06T21:15:51.9149132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 66%] 2024-08-06T21:15:51.9150355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 66%] 2024-08-06T21:15:51.9151601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 66%] 2024-08-06T21:15:51.9152865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 66%] 2024-08-06T21:15:51.9154078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 66%] 2024-08-06T21:15:51.9155277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 66%] 2024-08-06T21:15:51.9156509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 66%] 2024-08-06T21:15:51.9157714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 66%] 2024-08-06T21:15:51.9158932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 66%] 2024-08-06T21:15:51.9160133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 66%] 2024-08-06T21:15:51.9161393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 66%] 2024-08-06T21:15:51.9162653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 66%] 2024-08-06T21:15:51.9163857Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 66%] 2024-08-06T21:15:51.9165081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0050s] [ 66%] 2024-08-06T21:15:51.9166291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 66%] 2024-08-06T21:15:51.9167532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 66%] 2024-08-06T21:15:51.9168733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 66%] 2024-08-06T21:15:51.9169964Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 66%] 2024-08-06T21:15:51.9171215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 66%] 2024-08-06T21:15:51.9172486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 66%] 2024-08-06T21:15:51.9173682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 66%] 2024-08-06T21:15:51.9174889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 66%] 2024-08-06T21:15:51.9176100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 66%] 2024-08-06T21:15:51.9177310Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 66%] 2024-08-06T21:15:51.9178526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 66%] 2024-08-06T21:15:51.9179773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 66%] 2024-08-06T21:15:51.9181042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 66%] 2024-08-06T21:15:51.9182239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 67%] 2024-08-06T21:15:51.9183456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 67%] 2024-08-06T21:15:51.9184663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 67%] 2024-08-06T21:15:51.9185864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0047s] [ 67%] 2024-08-06T21:15:51.9187099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 67%] 2024-08-06T21:15:51.9188302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 67%] 2024-08-06T21:15:51.9189614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 67%] 2024-08-06T21:15:51.9190874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 67%] 2024-08-06T21:15:51.9192099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 67%] 2024-08-06T21:15:51.9193291Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 67%] 2024-08-06T21:15:51.9194508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 67%] 2024-08-06T21:15:51.9195698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 67%] 2024-08-06T21:15:51.9196912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 67%] 2024-08-06T21:15:51.9198170Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 67%] 2024-08-06T21:15:51.9199437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 67%] 2024-08-06T21:15:51.9200643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 67%] 2024-08-06T21:15:51.9201843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 67%] 2024-08-06T21:15:51.9203077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0085s] [ 67%] 2024-08-06T21:15:51.9204693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0086s] [ 67%] 2024-08-06T21:15:51.9205923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 67%] 2024-08-06T21:15:51.9207175Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 67%] 2024-08-06T21:15:51.9208465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0095s] [ 67%] 2024-08-06T21:15:51.9209776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0096s] [ 67%] 2024-08-06T21:15:51.9210992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 67%] 2024-08-06T21:15:51.9212234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 67%] 2024-08-06T21:15:51.9213443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 67%] 2024-08-06T21:15:51.9214673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 67%] 2024-08-06T21:15:51.9215874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 67%] 2024-08-06T21:15:51.9217159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 67%] 2024-08-06T21:15:51.9218446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 67%] 2024-08-06T21:15:51.9219668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 67%] 2024-08-06T21:15:51.9220891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 67%] 2024-08-06T21:15:51.9222108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 67%] 2024-08-06T21:15:51.9223342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0092s] [ 67%] 2024-08-06T21:15:51.9224560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0091s] [ 67%] 2024-08-06T21:15:51.9225796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 67%] 2024-08-06T21:15:51.9227080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 67%] 2024-08-06T21:15:51.9228349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0102s] [ 67%] 2024-08-06T21:15:51.9229589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0101s] [ 67%] 2024-08-06T21:15:51.9230805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 67%] 2024-08-06T21:15:51.9232049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 67%] 2024-08-06T21:15:51.9233256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 67%] 2024-08-06T21:15:51.9234482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 67%] 2024-08-06T21:15:51.9235726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 67%] 2024-08-06T21:15:51.9237025Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 67%] 2024-08-06T21:15:51.9238241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 67%] 2024-08-06T21:15:51.9239460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 67%] 2024-08-06T21:15:51.9240692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 67%] 2024-08-06T21:15:51.9241916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 67%] 2024-08-06T21:15:51.9243143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0110s] [ 67%] 2024-08-06T21:15:51.9244355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0111s] [ 67%] 2024-08-06T21:15:51.9245624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 67%] 2024-08-06T21:15:51.9246911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 67%] 2024-08-06T21:15:51.9248150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0119s] [ 67%] 2024-08-06T21:15:51.9249371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0116s] [ 67%] 2024-08-06T21:15:51.9250610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 67%] 2024-08-06T21:15:51.9251843Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 67%] 2024-08-06T21:15:51.9253055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 67%] 2024-08-06T21:15:51.9254323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 67%] 2024-08-06T21:15:51.9255579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 67%] 2024-08-06T21:15:51.9256828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 67%] 2024-08-06T21:15:51.9258038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 67%] 2024-08-06T21:15:51.9259284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 67%] 2024-08-06T21:15:51.9260500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 67%] 2024-08-06T21:15:51.9261737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 67%] 2024-08-06T21:15:51.9262942Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0078s] [ 67%] 2024-08-06T21:15:51.9264203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 67%] 2024-08-06T21:15:51.9265477Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 67%] 2024-08-06T21:15:51.9266702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 67%] 2024-08-06T21:15:51.9267932Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0091s] [ 67%] 2024-08-06T21:15:51.9269156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0092s] [ 67%] 2024-08-06T21:15:51.9270390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 67%] 2024-08-06T21:15:51.9271606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 67%] 2024-08-06T21:15:51.9272885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 67%] 2024-08-06T21:15:51.9274147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 67%] 2024-08-06T21:15:51.9275355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 67%] 2024-08-06T21:15:51.9276604Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 67%] 2024-08-06T21:15:51.9277813Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 67%] 2024-08-06T21:15:51.9279048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 68%] 2024-08-06T21:15:51.9280254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 68%] 2024-08-06T21:15:51.9281481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 68%] 2024-08-06T21:15:51.9282733Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 68%] 2024-08-06T21:15:51.9284063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 68%] 2024-08-06T21:15:51.9285268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 68%] 2024-08-06T21:15:51.9286478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0049s] [ 68%] 2024-08-06T21:15:51.9287713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 68%] 2024-08-06T21:15:51.9288938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 68%] 2024-08-06T21:15:51.9290166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 68%] 2024-08-06T21:15:51.9291425Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 68%] 2024-08-06T21:15:51.9292693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 68%] 2024-08-06T21:15:51.9293903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 68%] 2024-08-06T21:15:51.9295118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 68%] 2024-08-06T21:15:51.9296324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 68%] 2024-08-06T21:15:51.9297547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 68%] 2024-08-06T21:15:51.9298779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 68%] 2024-08-06T21:15:51.9299981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 68%] 2024-08-06T21:15:51.9301256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 68%] 2024-08-06T21:15:51.9302513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 68%] 2024-08-06T21:15:51.9303771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0049s] [ 68%] 2024-08-06T21:15:51.9305290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 68%] 2024-08-06T21:15:51.9306527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 68%] 2024-08-06T21:15:51.9307774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 68%] 2024-08-06T21:15:51.9308987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 68%] 2024-08-06T21:15:51.9310216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 68%] 2024-08-06T21:15:51.9311501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 68%] 2024-08-06T21:15:51.9312789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 68%] 2024-08-06T21:15:51.9313990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 68%] 2024-08-06T21:15:51.9315212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 68%] 2024-08-06T21:15:51.9316419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 68%] 2024-08-06T21:15:51.9317640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 68%] 2024-08-06T21:15:51.9318866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 68%] 2024-08-06T21:15:51.9320068Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 68%] 2024-08-06T21:15:51.9321353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 68%] 2024-08-06T21:15:51.9322619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 68%] 2024-08-06T21:15:51.9323845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 68%] 2024-08-06T21:15:51.9325050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 68%] 2024-08-06T21:15:51.9326279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 68%] 2024-08-06T21:15:51.9327505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 68%] 2024-08-06T21:15:51.9328720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 68%] 2024-08-06T21:15:51.9329989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 68%] 2024-08-06T21:15:51.9331257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 68%] 2024-08-06T21:15:51.9332471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 68%] 2024-08-06T21:15:51.9333677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 68%] 2024-08-06T21:15:51.9334903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 68%] 2024-08-06T21:15:51.9336108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 68%] 2024-08-06T21:15:51.9337348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 68%] 2024-08-06T21:15:51.9338556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 68%] 2024-08-06T21:15:51.9339801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 68%] 2024-08-06T21:15:51.9341089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 68%] 2024-08-06T21:15:51.9342292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 68%] 2024-08-06T21:15:51.9343516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 68%] 2024-08-06T21:15:51.9344722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 68%] 2024-08-06T21:15:51.9345941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 68%] 2024-08-06T21:15:51.9347162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0070s] [ 68%] 2024-08-06T21:15:51.9348433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 68%] 2024-08-06T21:15:51.9349689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 68%] 2024-08-06T21:15:51.9350901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 68%] 2024-08-06T21:15:51.9352104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 68%] 2024-08-06T21:15:51.9353303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 68%] 2024-08-06T21:15:51.9354516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 68%] 2024-08-06T21:15:51.9355710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 68%] 2024-08-06T21:15:51.9356945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 68%] 2024-08-06T21:15:51.9358193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 68%] 2024-08-06T21:15:51.9359456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 68%] 2024-08-06T21:15:51.9360657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 68%] 2024-08-06T21:15:51.9361865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 68%] 2024-08-06T21:15:51.9363098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 68%] 2024-08-06T21:15:51.9364300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 68%] 2024-08-06T21:15:51.9365518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 68%] 2024-08-06T21:15:51.9366789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 68%] 2024-08-06T21:15:51.9368074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 68%] 2024-08-06T21:15:51.9369279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 68%] 2024-08-06T21:15:51.9370502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 68%] 2024-08-06T21:15:51.9371701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 68%] 2024-08-06T21:15:51.9372909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 68%] 2024-08-06T21:15:51.9374117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 69%] 2024-08-06T21:15:51.9375314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 69%] 2024-08-06T21:15:51.9376587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 69%] 2024-08-06T21:15:51.9377850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 69%] 2024-08-06T21:15:51.9379069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 69%] 2024-08-06T21:15:51.9380272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 69%] 2024-08-06T21:15:51.9381505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 69%] 2024-08-06T21:15:51.9382716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 69%] 2024-08-06T21:15:51.9383919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 69%] 2024-08-06T21:15:51.9385180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 69%] 2024-08-06T21:15:51.9386454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 69%] 2024-08-06T21:15:51.9387690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 69%] 2024-08-06T21:15:51.9388895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 69%] 2024-08-06T21:15:51.9390131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 69%] 2024-08-06T21:15:51.9391336Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 69%] 2024-08-06T21:15:51.9392550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 69%] 2024-08-06T21:15:51.9393743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 69%] 2024-08-06T21:15:51.9394991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 69%] 2024-08-06T21:15:51.9396263Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 69%] 2024-08-06T21:15:51.9397480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 69%] 2024-08-06T21:15:51.9398698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 69%] 2024-08-06T21:15:51.9399903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 69%] 2024-08-06T21:15:51.9401129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 69%] 2024-08-06T21:15:51.9402333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 69%] 2024-08-06T21:15:51.9403586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 69%] 2024-08-06T21:15:51.9405193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 69%] 2024-08-06T21:15:51.9406426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 69%] 2024-08-06T21:15:51.9407670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 69%] 2024-08-06T21:15:51.9408878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 69%] 2024-08-06T21:15:51.9410108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 69%] 2024-08-06T21:15:51.9411295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 69%] 2024-08-06T21:15:51.9412502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 69%] 2024-08-06T21:15:51.9413780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 69%] 2024-08-06T21:15:51.9415070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 69%] 2024-08-06T21:15:51.9416272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 69%] 2024-08-06T21:15:51.9417487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 69%] 2024-08-06T21:15:51.9418708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 69%] 2024-08-06T21:15:51.9419919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 69%] 2024-08-06T21:15:51.9421137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 69%] 2024-08-06T21:15:51.9422335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 69%] 2024-08-06T21:15:51.9423609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 69%] 2024-08-06T21:15:51.9424877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 69%] 2024-08-06T21:15:51.9426099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 69%] 2024-08-06T21:15:51.9427325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 69%] 2024-08-06T21:15:51.9428530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 69%] 2024-08-06T21:15:51.9429759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 69%] 2024-08-06T21:15:51.9430946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 69%] 2024-08-06T21:15:51.9432206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 69%] 2024-08-06T21:15:51.9433459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 69%] 2024-08-06T21:15:51.9434671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 69%] 2024-08-06T21:15:51.9435861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 69%] 2024-08-06T21:15:51.9437110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 69%] 2024-08-06T21:15:51.9438314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 69%] 2024-08-06T21:15:51.9439513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_1_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 69%] 2024-08-06T21:15:51.9440755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0100s] [ 69%] 2024-08-06T21:15:51.9442026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 69%] 2024-08-06T21:15:51.9443324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0075s] [ 69%] 2024-08-06T21:15:51.9444547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 69%] 2024-08-06T21:15:51.9445794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0130s] [ 69%] 2024-08-06T21:15:51.9447055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0131s] [ 69%] 2024-08-06T21:15:51.9448298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0084s] [ 69%] 2024-08-06T21:15:51.9449527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0085s] [ 69%] 2024-08-06T21:15:51.9450745Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0098s] [ 69%] 2024-08-06T21:15:51.9452034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0100s] [ 69%] 2024-08-06T21:15:51.9453296Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0075s] [ 69%] 2024-08-06T21:15:51.9454533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0076s] [ 69%] 2024-08-06T21:15:51.9455761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0120s] [ 69%] 2024-08-06T21:15:51.9457035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0122s] [ 69%] 2024-08-06T21:15:51.9458256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0088s] [ 69%] 2024-08-06T21:15:51.9459501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0087s] [ 69%] 2024-08-06T21:15:51.9460767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0110s] [ 69%] 2024-08-06T21:15:51.9462048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0114s] [ 69%] 2024-08-06T21:15:51.9463284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0083s] [ 69%] 2024-08-06T21:15:51.9464513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 69%] 2024-08-06T21:15:51.9465761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0136s] [ 69%] 2024-08-06T21:15:51.9467018Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0138s] [ 69%] 2024-08-06T21:15:51.9468257Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0093s] [ 69%] 2024-08-06T21:15:51.9469532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0091s] [ 70%] 2024-08-06T21:15:51.9470817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0105s] [ 70%] 2024-08-06T21:15:51.9472039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0108s] [ 70%] 2024-08-06T21:15:51.9473256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0082s] [ 70%] 2024-08-06T21:15:51.9474502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0085s] [ 70%] 2024-08-06T21:15:51.9475727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0129s] [ 70%] 2024-08-06T21:15:51.9476998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0129s] [ 70%] 2024-08-06T21:15:51.9478217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0093s] [ 70%] 2024-08-06T21:15:51.9479500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0092s] [ 70%] 2024-08-06T21:15:51.9480798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0129s] [ 70%] 2024-08-06T21:15:51.9482048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0131s] [ 70%] 2024-08-06T21:15:51.9483273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0096s] [ 70%] 2024-08-06T21:15:51.9484507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0094s] [ 70%] 2024-08-06T21:15:51.9485758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0155s] [ 70%] 2024-08-06T21:15:51.9487009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0156s] [ 70%] 2024-08-06T21:15:51.9488300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0106s] [ 70%] 2024-08-06T21:15:51.9489621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0108s] [ 70%] 2024-08-06T21:15:51.9490869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0124s] [ 70%] 2024-08-06T21:15:51.9492093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0126s] [ 70%] 2024-08-06T21:15:51.9493332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0098s] [ 70%] 2024-08-06T21:15:51.9494559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0098s] [ 70%] 2024-08-06T21:15:51.9495780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0147s] [ 70%] 2024-08-06T21:15:51.9497044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0148s] [ 70%] 2024-08-06T21:15:51.9498307Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0108s] [ 70%] 2024-08-06T21:15:51.9499609Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0106s] [ 70%] 2024-08-06T21:15:51.9500829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0098s] [ 70%] 2024-08-06T21:15:51.9502073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0098s] [ 70%] 2024-08-06T21:15:51.9503299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0072s] [ 70%] 2024-08-06T21:15:51.9504995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 70%] 2024-08-06T21:15:51.9506241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0126s] [ 70%] 2024-08-06T21:15:51.9507572Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0126s] [ 70%] 2024-08-06T21:15:51.9508885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0080s] [ 70%] 2024-08-06T21:15:51.9510115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 70%] 2024-08-06T21:15:51.9511346Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0094s] [ 70%] 2024-08-06T21:15:51.9512570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0097s] [ 70%] 2024-08-06T21:15:51.9513803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0073s] [ 70%] 2024-08-06T21:15:51.9515350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 70%] 2024-08-06T21:15:51.9516836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0117s] [ 70%] 2024-08-06T21:15:51.9518224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0118s] [ 70%] 2024-08-06T21:15:51.9519544Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0083s] [ 70%] 2024-08-06T21:15:51.9520791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0083s] [ 70%] 2024-08-06T21:15:51.9522017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0165s] [ 70%] 2024-08-06T21:15:51.9523278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0174s] [ 70%] 2024-08-06T21:15:51.9524499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0110s] [ 70%] 2024-08-06T21:15:51.9525743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0111s] [ 70%] 2024-08-06T21:15:51.9527014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0218s] [ 70%] 2024-08-06T21:15:51.9528344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0221s] [ 70%] 2024-08-06T21:15:51.9529568Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0128s] [ 70%] 2024-08-06T21:15:51.9530819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0128s] [ 70%] 2024-08-06T21:15:51.9532035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0137s] [ 70%] 2024-08-06T21:15:51.9533259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0137s] [ 70%] 2024-08-06T21:15:51.9534489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0112s] [ 70%] 2024-08-06T21:15:51.9535706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0111s] [ 70%] 2024-08-06T21:15:51.9536987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0163s] [ 70%] 2024-08-06T21:15:51.9538286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0164s] [ 70%] 2024-08-06T21:15:51.9539527Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0125s] [ 70%] 2024-08-06T21:15:51.9540751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0127s] [ 70%] 2024-08-06T21:15:51.9541998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0181s] [ 70%] 2024-08-06T21:15:51.9543229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0190s] [ 70%] 2024-08-06T21:15:51.9544452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0126s] [ 70%] 2024-08-06T21:15:51.9545735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0123s] [ 70%] 2024-08-06T21:15:51.9547013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0232s] [ 70%] 2024-08-06T21:15:51.9548292Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0236s] [ 70%] 2024-08-06T21:15:51.9549520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0141s] [ 70%] 2024-08-06T21:15:51.9550768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0140s] [ 70%] 2024-08-06T21:15:51.9551992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0148s] [ 70%] 2024-08-06T21:15:51.9553226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0151s] [ 70%] 2024-08-06T21:15:51.9554439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0124s] [ 70%] 2024-08-06T21:15:51.9555713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0124s] [ 70%] 2024-08-06T21:15:51.9556997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0177s] [ 70%] 2024-08-06T21:15:51.9558250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0177s] [ 70%] 2024-08-06T21:15:51.9559492Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0140s] [ 70%] 2024-08-06T21:15:51.9560723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0142s] [ 70%] 2024-08-06T21:15:51.9561962Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0212s] [ 70%] 2024-08-06T21:15:51.9563188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0222s] [ 70%] 2024-08-06T21:15:51.9564469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.1863s] [ 70%] 2024-08-06T21:15:51.9565748Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0153s] [ 70%] 2024-08-06T21:15:51.9566972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0270s] [ 70%] 2024-08-06T21:15:51.9568243Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0274s] [ 71%] 2024-08-06T21:15:51.9569471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0170s] [ 71%] 2024-08-06T21:15:51.9570723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0169s] [ 71%] 2024-08-06T21:15:51.9571938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0178s] [ 71%] 2024-08-06T21:15:51.9573217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0181s] [ 71%] 2024-08-06T21:15:51.9574485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0151s] [ 71%] 2024-08-06T21:15:51.9575724Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0152s] [ 71%] 2024-08-06T21:15:51.9576946Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0209s] [ 71%] 2024-08-06T21:15:51.9578191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0208s] [ 71%] 2024-08-06T21:15:51.9579442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0171s] [ 71%] 2024-08-06T21:15:51.9580669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0169s] [ 71%] 2024-08-06T21:15:51.9581904Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0158s] [ 71%] 2024-08-06T21:15:51.9583214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0163s] [ 71%] 2024-08-06T21:15:51.9584507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0107s] [ 71%] 2024-08-06T21:15:51.9585727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0108s] [ 71%] 2024-08-06T21:15:51.9586972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0216s] [ 71%] 2024-08-06T21:15:51.9588223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0217s] [ 71%] 2024-08-06T21:15:51.9589455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0127s] [ 71%] 2024-08-06T21:15:51.9590702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0127s] [ 71%] 2024-08-06T21:15:51.9591952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0131s] [ 71%] 2024-08-06T21:15:51.9593242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0131s] [ 71%] 2024-08-06T21:15:51.9594458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0109s] [ 71%] 2024-08-06T21:15:51.9595698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0109s] [ 71%] 2024-08-06T21:15:51.9596919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0160s] [ 71%] 2024-08-06T21:15:51.9598190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0159s] [ 71%] 2024-08-06T21:15:51.9599410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0134s] [ 71%] 2024-08-06T21:15:51.9600630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0128s] [ 71%] 2024-08-06T21:15:51.9601906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0064s] [ 71%] 2024-08-06T21:15:51.9603180Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 71%] 2024-08-06T21:15:51.9604859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 71%] 2024-08-06T21:15:51.9606105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 71%] 2024-08-06T21:15:51.9607349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 71%] 2024-08-06T21:15:51.9608602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 71%] 2024-08-06T21:15:51.9609838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 71%] 2024-08-06T21:15:51.9611131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 71%] 2024-08-06T21:15:51.9612410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0064s] [ 71%] 2024-08-06T21:15:51.9613642Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 71%] 2024-08-06T21:15:51.9614846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 71%] 2024-08-06T21:15:51.9616081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 71%] 2024-08-06T21:15:51.9617797Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0076s] [ 71%] 2024-08-06T21:15:51.9619088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 71%] 2024-08-06T21:15:51.9620300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 71%] 2024-08-06T21:15:51.9621623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 71%] 2024-08-06T21:15:51.9622924Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0062s] [ 71%] 2024-08-06T21:15:51.9624145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 71%] 2024-08-06T21:15:51.9625376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 71%] 2024-08-06T21:15:51.9626597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 71%] 2024-08-06T21:15:51.9627838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 71%] 2024-08-06T21:15:51.9629084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 71%] 2024-08-06T21:15:51.9630318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 71%] 2024-08-06T21:15:51.9631584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 71%] 2024-08-06T21:15:51.9632855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 71%] 2024-08-06T21:15:51.9634064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 71%] 2024-08-06T21:15:51.9635268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0062s] [ 71%] 2024-08-06T21:15:51.9636503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 71%] 2024-08-06T21:15:51.9637715Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0080s] [ 71%] 2024-08-06T21:15:51.9638972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 71%] 2024-08-06T21:15:51.9640250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 71%] 2024-08-06T21:15:51.9641543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 71%] 2024-08-06T21:15:51.9642755Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0072s] [ 71%] 2024-08-06T21:15:51.9643992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 71%] 2024-08-06T21:15:51.9645208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 71%] 2024-08-06T21:15:51.9646434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 71%] 2024-08-06T21:15:51.9647671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0083s] [ 71%] 2024-08-06T21:15:51.9648914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0085s] [ 71%] 2024-08-06T21:15:51.9650186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 71%] 2024-08-06T21:15:51.9651459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 71%] 2024-08-06T21:15:51.9652684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0074s] [ 71%] 2024-08-06T21:15:51.9653897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 71%] 2024-08-06T21:15:51.9655123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 71%] 2024-08-06T21:15:51.9656339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 71%] 2024-08-06T21:15:51.9657546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0086s] [ 71%] 2024-08-06T21:15:51.9658837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 71%] 2024-08-06T21:15:51.9660617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 71%] 2024-08-06T21:15:51.9662822Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 71%] 2024-08-06T21:15:51.9664056Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 71%] 2024-08-06T21:15:51.9665299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 71%] 2024-08-06T21:15:51.9666519Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 72%] 2024-08-06T21:15:51.9667754Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 72%] 2024-08-06T21:15:51.9668966Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0072s] [ 72%] 2024-08-06T21:15:51.9670271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 72%] 2024-08-06T21:15:51.9672204Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 72%] 2024-08-06T21:15:51.9673713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 72%] 2024-08-06T21:15:51.9674969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 72%] 2024-08-06T21:15:51.9676188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 72%] 2024-08-06T21:15:51.9677411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 72%] 2024-08-06T21:15:51.9678615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 72%] 2024-08-06T21:15:51.9680126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0074s] [ 72%] 2024-08-06T21:15:51.9681834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 72%] 2024-08-06T21:15:51.9683325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 72%] 2024-08-06T21:15:51.9685621Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 72%] 2024-08-06T21:15:51.9688681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0294s] [ 72%] 2024-08-06T21:15:51.9689995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.1986s] [ 72%] 2024-08-06T21:15:51.9691610Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0185s] [ 72%] 2024-08-06T21:15:51.9693063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0184s] [ 72%] 2024-08-06T21:15:51.9695259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0403s] [ 72%] 2024-08-06T21:15:51.9697591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0416s] [ 72%] 2024-08-06T21:15:51.9700373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0218s] [ 72%] 2024-08-06T21:15:51.9703129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0218s] [ 72%] 2024-08-06T21:15:51.9706324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0211s] [ 72%] 2024-08-06T21:15:51.9708372Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0213s] [ 72%] 2024-08-06T21:15:51.9710765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0187s] [ 72%] 2024-08-06T21:15:51.9712044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0186s] [ 72%] 2024-08-06T21:15:51.9713434Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0258s] [ 72%] 2024-08-06T21:15:51.9716637Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0257s] [ 72%] 2024-08-06T21:15:51.9718227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0217s] [ 72%] 2024-08-06T21:15:51.9720433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0219s] [ 72%] 2024-08-06T21:15:51.9722279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0330s] [ 72%] 2024-08-06T21:15:51.9723538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0349s] [ 72%] 2024-08-06T21:15:51.9725803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0211s] [ 72%] 2024-08-06T21:15:51.9727216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0212s] [ 72%] 2024-08-06T21:15:51.9728819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0434s] [ 72%] 2024-08-06T21:15:51.9730405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0441s] [ 72%] 2024-08-06T21:15:51.9731676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.1862s] [ 72%] 2024-08-06T21:15:51.9732935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0244s] [ 72%] 2024-08-06T21:15:51.9734442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0237s] [ 72%] 2024-08-06T21:15:51.9736160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0236s] [ 72%] 2024-08-06T21:15:51.9737488Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0210s] [ 72%] 2024-08-06T21:15:51.9738771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0209s] [ 72%] 2024-08-06T21:15:51.9740535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0279s] [ 72%] 2024-08-06T21:15:51.9742035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0280s] [ 72%] 2024-08-06T21:15:51.9743284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0239s] [ 72%] 2024-08-06T21:15:51.9744516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0238s] [ 72%] 2024-08-06T21:15:51.9745753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0384s] [ 72%] 2024-08-06T21:15:51.9746980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0404s] [ 72%] 2024-08-06T21:15:51.9748273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0256s] [ 72%] 2024-08-06T21:15:51.9749557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0254s] [ 72%] 2024-08-06T21:15:51.9750786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0488s] [ 72%] 2024-08-06T21:15:51.9752046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0495s] [ 72%] 2024-08-06T21:15:51.9753279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0291s] [ 72%] 2024-08-06T21:15:51.9754530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.1931s] [ 72%] 2024-08-06T21:15:51.9755741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0273s] [ 72%] 2024-08-06T21:15:51.9757029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0276s] [ 72%] 2024-08-06T21:15:51.9758300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0247s] [ 72%] 2024-08-06T21:15:51.9759542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0250s] [ 72%] 2024-08-06T21:15:51.9760761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0324s] [ 72%] 2024-08-06T21:15:51.9762001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0324s] [ 72%] 2024-08-06T21:15:51.9763250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0281s] [ 72%] 2024-08-06T21:15:51.9764471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0279s] [ 72%] 2024-08-06T21:15:51.9765699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0286s] [ 72%] 2024-08-06T21:15:51.9766969Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0295s] [ 72%] 2024-08-06T21:15:51.9768261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0176s] [ 72%] 2024-08-06T21:15:51.9769485Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0177s] [ 72%] 2024-08-06T21:15:51.9770727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0390s] [ 72%] 2024-08-06T21:15:51.9771975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0396s] [ 72%] 2024-08-06T21:15:51.9773201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0204s] [ 72%] 2024-08-06T21:15:51.9774444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0207s] [ 72%] 2024-08-06T21:15:51.9775699Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0198s] [ 72%] 2024-08-06T21:15:51.9776990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0201s] [ 72%] 2024-08-06T21:15:51.9778201Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0177s] [ 72%] 2024-08-06T21:15:51.9779441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.1853s] [ 72%] 2024-08-06T21:15:51.9780663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0245s] [ 72%] 2024-08-06T21:15:51.9781919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0244s] [ 72%] 2024-08-06T21:15:51.9783132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0207s] [ 72%] 2024-08-06T21:15:51.9784352Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0204s] [ 72%] 2024-08-06T21:15:51.9785629Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0071s] [ 73%] 2024-08-06T21:15:51.9786919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 73%] 2024-08-06T21:15:51.9788150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0067s] [ 73%] 2024-08-06T21:15:51.9789373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 73%] 2024-08-06T21:15:51.9790675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0090s] [ 73%] 2024-08-06T21:15:51.9791916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0091s] [ 73%] 2024-08-06T21:15:51.9793151Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0077s] [ 73%] 2024-08-06T21:15:51.9794420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0077s] [ 73%] 2024-08-06T21:15:51.9795687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0074s] [ 73%] 2024-08-06T21:15:51.9796919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 73%] 2024-08-06T21:15:51.9798121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 73%] 2024-08-06T21:15:51.9799358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 73%] 2024-08-06T21:15:51.9800573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0091s] [ 73%] 2024-08-06T21:15:51.9801811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0095s] [ 73%] 2024-08-06T21:15:51.9803022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0078s] [ 73%] 2024-08-06T21:15:51.9804805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 73%] 2024-08-06T21:15:51.9806812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0079s] [ 73%] 2024-08-06T21:15:51.9808039Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 73%] 2024-08-06T21:15:51.9809273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0074s] [ 73%] 2024-08-06T21:15:51.9810500Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 73%] 2024-08-06T21:15:51.9811739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0098s] [ 73%] 2024-08-06T21:15:51.9812968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0099s] [ 73%] 2024-08-06T21:15:51.9814261Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0083s] [ 73%] 2024-08-06T21:15:51.9815552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0083s] [ 73%] 2024-08-06T21:15:51.9816775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0081s] [ 73%] 2024-08-06T21:15:51.9817986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 73%] 2024-08-06T21:15:51.9819197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0076s] [ 73%] 2024-08-06T21:15:51.9820436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 73%] 2024-08-06T21:15:51.9821650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0091s] [ 73%] 2024-08-06T21:15:51.9822892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0092s] [ 73%] 2024-08-06T21:15:51.9824147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0075s] [ 73%] 2024-08-06T21:15:51.9825442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 73%] 2024-08-06T21:15:51.9826657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0085s] [ 73%] 2024-08-06T21:15:51.9827898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0088s] [ 73%] 2024-08-06T21:15:51.9829120Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 73%] 2024-08-06T21:15:51.9830344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 73%] 2024-08-06T21:15:51.9831577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0098s] [ 73%] 2024-08-06T21:15:51.9832848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0097s] [ 73%] 2024-08-06T21:15:51.9834142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0078s] [ 73%] 2024-08-06T21:15:51.9835368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 73%] 2024-08-06T21:15:51.9836588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0087s] [ 73%] 2024-08-06T21:15:51.9837806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0091s] [ 73%] 2024-08-06T21:15:51.9839036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0074s] [ 73%] 2024-08-06T21:15:51.9840249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 73%] 2024-08-06T21:15:51.9841461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0104s] [ 73%] 2024-08-06T21:15:51.9842768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0104s] [ 73%] 2024-08-06T21:15:51.9844040Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0083s] [ 73%] 2024-08-06T21:15:51.9845283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0081s] [ 73%] 2024-08-06T21:15:51.9846495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0067s] [ 73%] 2024-08-06T21:15:51.9847742Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 73%] 2024-08-06T21:15:51.9848954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 73%] 2024-08-06T21:15:51.9850186Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 73%] 2024-08-06T21:15:51.9851446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0080s] [ 73%] 2024-08-06T21:15:51.9852739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 73%] 2024-08-06T21:15:51.9853974Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 73%] 2024-08-06T21:15:51.9855192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 73%] 2024-08-06T21:15:51.9856423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0072s] [ 73%] 2024-08-06T21:15:51.9857636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 73%] 2024-08-06T21:15:51.9858853Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0058s] [ 73%] 2024-08-06T21:15:51.9860061Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 73%] 2024-08-06T21:15:51.9861343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0083s] [ 73%] 2024-08-06T21:15:51.9862607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 73%] 2024-08-06T21:15:51.9863816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 73%] 2024-08-06T21:15:51.9865051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 73%] 2024-08-06T21:15:51.9866269Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 73%] 2024-08-06T21:15:51.9867513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 73%] 2024-08-06T21:15:51.9868721Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 73%] 2024-08-06T21:15:51.9869957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 73%] 2024-08-06T21:15:51.9871218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0075s] [ 73%] 2024-08-06T21:15:51.9872503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0074s] [ 73%] 2024-08-06T21:15:51.9873726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 73%] 2024-08-06T21:15:51.9874949Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 73%] 2024-08-06T21:15:51.9876182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 73%] 2024-08-06T21:15:51.9877386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 73%] 2024-08-06T21:15:51.9878605Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 73%] 2024-08-06T21:15:51.9879858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 73%] 2024-08-06T21:15:51.9881137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0077s] [ 73%] 2024-08-06T21:15:51.9882347Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 74%] 2024-08-06T21:15:51.9883583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 74%] 2024-08-06T21:15:51.9884807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 74%] 2024-08-06T21:15:51.9886021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 74%] 2024-08-06T21:15:51.9887251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 74%] 2024-08-06T21:15:51.9888467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 74%] 2024-08-06T21:15:51.9889741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 74%] 2024-08-06T21:15:51.9891006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0079s] [ 74%] 2024-08-06T21:15:51.9892244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 74%] 2024-08-06T21:15:51.9893474Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 74%] 2024-08-06T21:15:51.9894718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 74%] 2024-08-06T21:15:51.9895922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 74%] 2024-08-06T21:15:51.9897131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 74%] 2024-08-06T21:15:51.9898395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 74%] 2024-08-06T21:15:51.9899658Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 74%] 2024-08-06T21:15:51.9900882Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0081s] [ 74%] 2024-08-06T21:15:51.9902092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 74%] 2024-08-06T21:15:51.9903331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 74%] 2024-08-06T21:15:51.9905015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 74%] 2024-08-06T21:15:51.9906278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0075s] [ 74%] 2024-08-06T21:15:51.9907499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 74%] 2024-08-06T21:15:51.9908795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0067s] [ 74%] 2024-08-06T21:15:51.9910098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 74%] 2024-08-06T21:15:51.9911314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0085s] [ 74%] 2024-08-06T21:15:51.9912556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0085s] [ 74%] 2024-08-06T21:15:51.9913795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 74%] 2024-08-06T21:15:51.9915032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 74%] 2024-08-06T21:15:51.9916238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0075s] [ 74%] 2024-08-06T21:15:51.9917522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 74%] 2024-08-06T21:15:51.9919328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0063s] [ 74%] 2024-08-06T21:15:51.9920573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 74%] 2024-08-06T21:15:51.9921800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0089s] [ 74%] 2024-08-06T21:15:51.9923017Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0088s] [ 74%] 2024-08-06T21:15:51.9924267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0072s] [ 74%] 2024-08-06T21:15:51.9925481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 74%] 2024-08-06T21:15:51.9926708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 74%] 2024-08-06T21:15:51.9927990Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 74%] 2024-08-06T21:15:51.9929289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 74%] 2024-08-06T21:15:51.9930504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 74%] 2024-08-06T21:15:51.9931726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0072s] [ 74%] 2024-08-06T21:15:51.9932972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 74%] 2024-08-06T21:15:51.9934210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 74%] 2024-08-06T21:15:51.9935445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 74%] 2024-08-06T21:15:51.9936690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 74%] 2024-08-06T21:15:51.9937967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 74%] 2024-08-06T21:15:51.9939169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 74%] 2024-08-06T21:15:51.9940396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 74%] 2024-08-06T21:15:51.9941606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0076s] [ 74%] 2024-08-06T21:15:51.9942824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 74%] 2024-08-06T21:15:51.9944071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 74%] 2024-08-06T21:15:51.9945282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_128_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 74%] 2024-08-06T21:15:51.9946570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0160s] [ 74%] 2024-08-06T21:15:51.9947848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0166s] [ 74%] 2024-08-06T21:15:51.9949088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0111s] [ 74%] 2024-08-06T21:15:51.9950314Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0109s] [ 74%] 2024-08-06T21:15:51.9951557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0215s] [ 74%] 2024-08-06T21:15:51.9952788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0217s] [ 74%] 2024-08-06T21:15:51.9954035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0126s] [ 74%] 2024-08-06T21:15:51.9955323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0126s] [ 74%] 2024-08-06T21:15:51.9956592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0159s] [ 74%] 2024-08-06T21:15:51.9957826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0162s] [ 74%] 2024-08-06T21:15:51.9959034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0114s] [ 74%] 2024-08-06T21:15:51.9960270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0111s] [ 74%] 2024-08-06T21:15:51.9961499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0213s] [ 74%] 2024-08-06T21:15:51.9962730Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0214s] [ 74%] 2024-08-06T21:15:51.9963972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0129s] [ 74%] 2024-08-06T21:15:51.9965256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0128s] [ 74%] 2024-08-06T21:15:51.9966524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0173s] [ 74%] 2024-08-06T21:15:51.9967753Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0183s] [ 74%] 2024-08-06T21:15:51.9968998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0128s] [ 74%] 2024-08-06T21:15:51.9970232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0125s] [ 74%] 2024-08-06T21:15:51.9971468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0229s] [ 74%] 2024-08-06T21:15:51.9972693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0231s] [ 74%] 2024-08-06T21:15:51.9974014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0143s] [ 74%] 2024-08-06T21:15:51.9975293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0141s] [ 74%] 2024-08-06T21:15:51.9976524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0175s] [ 74%] 2024-08-06T21:15:51.9977734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.1910s] [ 74%] 2024-08-06T21:15:51.9985000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0137s] [ 75%] 2024-08-06T21:15:51.9986356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0135s] [ 75%] 2024-08-06T21:15:51.9987583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0230s] [ 75%] 2024-08-06T21:15:51.9988821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0233s] [ 75%] 2024-08-06T21:15:51.9990156Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0152s] [ 75%] 2024-08-06T21:15:51.9991453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0149s] [ 75%] 2024-08-06T21:15:51.9992667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0212s] [ 75%] 2024-08-06T21:15:51.9993900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0220s] [ 75%] 2024-08-06T21:15:51.9995122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0154s] [ 75%] 2024-08-06T21:15:51.9996348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0156s] [ 75%] 2024-08-06T21:15:51.9997564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0265s] [ 75%] 2024-08-06T21:15:51.9998838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0265s] [ 75%] 2024-08-06T21:15:52.0000124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0170s] [ 75%] 2024-08-06T21:15:52.0001349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0172s] [ 75%] 2024-08-06T21:15:52.0002569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0211s] [ 75%] 2024-08-06T21:15:52.0003785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0216s] [ 75%] 2024-08-06T21:15:52.0005562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0155s] [ 75%] 2024-08-06T21:15:52.0006798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0157s] [ 75%] 2024-08-06T21:15:52.0008019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0262s] [ 75%] 2024-08-06T21:15:52.0009365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0264s] [ 75%] 2024-08-06T21:15:52.0010684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0174s] [ 75%] 2024-08-06T21:15:52.0011898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0174s] [ 75%] 2024-08-06T21:15:52.0013116Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0155s] [ 75%] 2024-08-06T21:15:52.0014343Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0157s] [ 75%] 2024-08-06T21:15:52.0015554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0110s] [ 75%] 2024-08-06T21:15:52.0016775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0109s] [ 75%] 2024-08-06T21:15:52.0018050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0209s] [ 75%] 2024-08-06T21:15:52.0019345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0211s] [ 75%] 2024-08-06T21:15:52.0020574Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0126s] [ 75%] 2024-08-06T21:15:52.0022236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.1824s] [ 75%] 2024-08-06T21:15:52.0023449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0154s] [ 75%] 2024-08-06T21:15:52.0024664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0156s] [ 75%] 2024-08-06T21:15:52.0025897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0112s] [ 75%] 2024-08-06T21:15:52.0027140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0113s] [ 75%] 2024-08-06T21:15:52.0028461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0204s] [ 75%] 2024-08-06T21:15:52.0029784Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0206s] [ 75%] 2024-08-06T21:15:52.0031015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0128s] [ 75%] 2024-08-06T21:15:52.0032228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0128s] [ 75%] 2024-08-06T21:15:52.0033456Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0287s] [ 75%] 2024-08-06T21:15:52.0034671Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0299s] [ 75%] 2024-08-06T21:15:52.0035883Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0186s] [ 75%] 2024-08-06T21:15:52.0037160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0182s] [ 75%] 2024-08-06T21:15:52.0038430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0397s] [ 75%] 2024-08-06T21:15:52.0039662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0398s] [ 75%] 2024-08-06T21:15:52.0040895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0216s] [ 75%] 2024-08-06T21:15:52.0042125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0216s] [ 75%] 2024-08-06T21:15:52.0043338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0255s] [ 75%] 2024-08-06T21:15:52.0044556Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0258s] [ 75%] 2024-08-06T21:15:52.0045767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0187s] [ 75%] 2024-08-06T21:15:52.0047022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0191s] [ 75%] 2024-08-06T21:15:52.0048284Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0342s] [ 75%] 2024-08-06T21:15:52.0049496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0347s] [ 75%] 2024-08-06T21:15:52.0050739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0214s] [ 75%] 2024-08-06T21:15:52.0051957Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0216s] [ 75%] 2024-08-06T21:15:52.0053174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0314s] [ 75%] 2024-08-06T21:15:52.0054390Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0323s] [ 75%] 2024-08-06T21:15:52.0055657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.1882s] [ 75%] 2024-08-06T21:15:52.0056931Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0208s] [ 75%] 2024-08-06T21:15:52.0058148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0417s] [ 75%] 2024-08-06T21:15:52.0059378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0424s] [ 75%] 2024-08-06T21:15:52.0060619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0241s] [ 75%] 2024-08-06T21:15:52.0061849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0241s] [ 75%] 2024-08-06T21:15:52.0063052Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0284s] [ 75%] 2024-08-06T21:15:52.0064273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0288s] [ 75%] 2024-08-06T21:15:52.0065521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0210s] [ 75%] 2024-08-06T21:15:52.0066790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0209s] [ 75%] 2024-08-06T21:15:52.0068005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0365s] [ 75%] 2024-08-06T21:15:52.0069227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0368s] [ 75%] 2024-08-06T21:15:52.0070478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0242s] [ 75%] 2024-08-06T21:15:52.0071690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0243s] [ 75%] 2024-08-06T21:15:52.0072912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0363s] [ 75%] 2024-08-06T21:15:52.0074169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0386s] [ 75%] 2024-08-06T21:15:52.0075459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0245s] [ 75%] 2024-08-06T21:15:52.0076676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0244s] [ 75%] 2024-08-06T21:15:52.0077901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0472s] [ 75%] 2024-08-06T21:15:52.0079129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0472s] [ 75%] 2024-08-06T21:15:52.0080366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.1959s] [ 75%] 2024-08-06T21:15:52.0081590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0280s] [ 75%] 2024-08-06T21:15:52.0082858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0334s] [ 76%] 2024-08-06T21:15:52.0084136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0344s] [ 76%] 2024-08-06T21:15:52.0085339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0246s] [ 76%] 2024-08-06T21:15:52.0086552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0246s] [ 76%] 2024-08-06T21:15:52.0087768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0410s] [ 76%] 2024-08-06T21:15:52.0089037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0414s] [ 76%] 2024-08-06T21:15:52.0090325Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0281s] [ 76%] 2024-08-06T21:15:52.0091545Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0282s] [ 76%] 2024-08-06T21:15:52.0092808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0268s] [ 76%] 2024-08-06T21:15:52.0094075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0278s] [ 76%] 2024-08-06T21:15:52.0095283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0171s] [ 76%] 2024-08-06T21:15:52.0096498Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0168s] [ 76%] 2024-08-06T21:15:52.0097728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0377s] [ 76%] 2024-08-06T21:15:52.0098952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0387s] [ 76%] 2024-08-06T21:15:52.0100194Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0204s] [ 76%] 2024-08-06T21:15:52.0101458Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0203s] [ 76%] 2024-08-06T21:15:52.0102714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0241s] [ 76%] 2024-08-06T21:15:52.0104008Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0244s] [ 76%] 2024-08-06T21:15:52.0105670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.1828s] [ 76%] 2024-08-06T21:15:52.0106916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0180s] [ 76%] 2024-08-06T21:15:52.0108137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0329s] [ 76%] 2024-08-06T21:15:52.0109357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0332s] [ 76%] 2024-08-06T21:15:52.0110579Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0207s] [ 76%] 2024-08-06T21:15:52.0111892Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0209s] [ 76%] 2024-08-06T21:15:52.0113172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0079s] [ 76%] 2024-08-06T21:15:52.0114381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0081s] [ 76%] 2024-08-06T21:15:52.0115591Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0071s] [ 76%] 2024-08-06T21:15:52.0116807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 76%] 2024-08-06T21:15:52.0118019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0106s] [ 76%] 2024-08-06T21:15:52.0119229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 76%] 2024-08-06T21:15:52.0120463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0077s] [ 76%] 2024-08-06T21:15:52.0121732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 76%] 2024-08-06T21:15:52.0123009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0088s] [ 76%] 2024-08-06T21:15:52.0124211Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 76%] 2024-08-06T21:15:52.0125409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0074s] [ 76%] 2024-08-06T21:15:52.0126630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 76%] 2024-08-06T21:15:52.0127834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0106s] [ 76%] 2024-08-06T21:15:52.0129051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0107s] [ 76%] 2024-08-06T21:15:52.0130322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0080s] [ 76%] 2024-08-06T21:15:52.0131597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 76%] 2024-08-06T21:15:52.0132800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0092s] [ 76%] 2024-08-06T21:15:52.0134019Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0090s] [ 76%] 2024-08-06T21:15:52.0135228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0075s] [ 76%] 2024-08-06T21:15:52.0136444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 76%] 2024-08-06T21:15:52.0137656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0114s] [ 76%] 2024-08-06T21:15:52.0138873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0113s] [ 76%] 2024-08-06T21:15:52.0140153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0084s] [ 76%] 2024-08-06T21:15:52.0141414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0083s] [ 76%] 2024-08-06T21:15:52.0142615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0091s] [ 76%] 2024-08-06T21:15:52.0143821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0092s] [ 76%] 2024-08-06T21:15:52.0145029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0068s] [ 76%] 2024-08-06T21:15:52.0146229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 76%] 2024-08-06T21:15:52.0147430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0109s] [ 76%] 2024-08-06T21:15:52.0148713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0110s] [ 76%] 2024-08-06T21:15:52.0149999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0076s] [ 76%] 2024-08-06T21:15:52.0151213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 76%] 2024-08-06T21:15:52.0152413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0110s] [ 76%] 2024-08-06T21:15:52.0153632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0110s] [ 76%] 2024-08-06T21:15:52.0154846Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0084s] [ 76%] 2024-08-06T21:15:52.0156057Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0081s] [ 76%] 2024-08-06T21:15:52.0157266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0124s] [ 76%] 2024-08-06T21:15:52.0158533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0124s] [ 76%] 2024-08-06T21:15:52.0159805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0087s] [ 76%] 2024-08-06T21:15:52.0161022Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0087s] [ 76%] 2024-08-06T21:15:52.0162232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0108s] [ 76%] 2024-08-06T21:15:52.0163443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0109s] [ 76%] 2024-08-06T21:15:52.0164649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0084s] [ 76%] 2024-08-06T21:15:52.0165850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0081s] [ 76%] 2024-08-06T21:15:52.0167105Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0126s] [ 76%] 2024-08-06T21:15:52.0168368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0128s] [ 76%] 2024-08-06T21:15:52.0169600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0087s] [ 76%] 2024-08-06T21:15:52.0170805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0088s] [ 76%] 2024-08-06T21:15:52.0172007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0074s] [ 76%] 2024-08-06T21:15:52.0173229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0073s] [ 76%] 2024-08-06T21:15:52.0174426Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 76%] 2024-08-06T21:15:52.0175644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 76%] 2024-08-06T21:15:52.0176894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0095s] [ 76%] 2024-08-06T21:15:52.0178169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0097s] [ 77%] 2024-08-06T21:15:52.0179386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 77%] 2024-08-06T21:15:52.0180623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 77%] 2024-08-06T21:15:52.0181830Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0076s] [ 77%] 2024-08-06T21:15:52.0183030Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0076s] [ 77%] 2024-08-06T21:15:52.0184232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 77%] 2024-08-06T21:15:52.0185473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 77%] 2024-08-06T21:15:52.0186735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0097s] [ 77%] 2024-08-06T21:15:52.0187978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0098s] [ 77%] 2024-08-06T21:15:52.0189209Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 77%] 2024-08-06T21:15:52.0190424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 77%] 2024-08-06T21:15:52.0191652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0531s] [ 77%] 2024-08-06T21:15:52.0192872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0560s] [ 77%] 2024-08-06T21:15:52.0194082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0316s] [ 77%] 2024-08-06T21:15:52.0195361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0316s] [ 77%] 2024-08-06T21:15:52.0196628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0719s] [ 77%] 2024-08-06T21:15:52.0197865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.2378s] [ 77%] 2024-08-06T21:15:52.0199102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0376s] [ 77%] 2024-08-06T21:15:52.0200348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0369s] [ 77%] 2024-08-06T21:15:52.0201554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0384s] [ 77%] 2024-08-06T21:15:52.0202765Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0388s] [ 77%] 2024-08-06T21:15:52.0204163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0316s] [ 77%] 2024-08-06T21:15:52.0205977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0317s] [ 77%] 2024-08-06T21:15:52.0207683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0488s] [ 77%] 2024-08-06T21:15:52.0209427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0491s] [ 77%] 2024-08-06T21:15:52.0211059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0370s] [ 77%] 2024-08-06T21:15:52.0213158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.1955s] [ 77%] 2024-08-06T21:15:52.0215430Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0560s] [ 77%] 2024-08-06T21:15:52.0216965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0597s] [ 77%] 2024-08-06T21:15:52.0219058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0358s] [ 77%] 2024-08-06T21:15:52.0220728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0342s] [ 77%] 2024-08-06T21:15:52.0222235Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0766s] [ 77%] 2024-08-06T21:15:52.0223663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0781s] [ 77%] 2024-08-06T21:15:52.0225094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0408s] [ 77%] 2024-08-06T21:15:52.0226350Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0408s] [ 77%] 2024-08-06T21:15:52.0228063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.1980s] [ 77%] 2024-08-06T21:15:52.0229763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0422s] [ 77%] 2024-08-06T21:15:52.0231912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0339s] [ 77%] 2024-08-06T21:15:52.0234012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0338s] [ 77%] 2024-08-06T21:15:52.0235612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0516s] [ 77%] 2024-08-06T21:15:52.0238139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0518s] [ 77%] 2024-08-06T21:15:52.0240293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0399s] [ 77%] 2024-08-06T21:15:52.0242301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0398s] [ 77%] 2024-08-06T21:15:52.0244375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0655s] [ 77%] 2024-08-06T21:15:52.0246767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.2228s] [ 77%] 2024-08-06T21:15:52.0248877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0412s] [ 77%] 2024-08-06T21:15:52.0250124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0410s] [ 77%] 2024-08-06T21:15:52.0252238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0848s] [ 77%] 2024-08-06T21:15:52.0253965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0867s] [ 77%] 2024-08-06T21:15:52.0255224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0497s] [ 77%] 2024-08-06T21:15:52.0257312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0494s] [ 77%] 2024-08-06T21:15:52.0258866Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0500s] [ 77%] 2024-08-06T21:15:52.0260174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.2116s] [ 77%] 2024-08-06T21:15:52.0261651Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0403s] [ 77%] 2024-08-06T21:15:52.0263739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0402s] [ 77%] 2024-08-06T21:15:52.0265264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0592s] [ 77%] 2024-08-06T21:15:52.0266704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0592s] [ 77%] 2024-08-06T21:15:52.0267937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0463s] [ 77%] 2024-08-06T21:15:52.0269157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0465s] [ 77%] 2024-08-06T21:15:52.0270440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0490s] [ 77%] 2024-08-06T21:15:52.0271712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.2083s] [ 77%] 2024-08-06T21:15:52.0272939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0288s] [ 77%] 2024-08-06T21:15:52.0274157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0287s] [ 77%] 2024-08-06T21:15:52.0275400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0697s] [ 77%] 2024-08-06T21:15:52.0276623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0713s] [ 77%] 2024-08-06T21:15:52.0277872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0346s] [ 77%] 2024-08-06T21:15:52.0279142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0346s] [ 77%] 2024-08-06T21:15:52.0280411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0358s] [ 77%] 2024-08-06T21:15:52.0281641Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0790s] [ 77%] 2024-08-06T21:15:52.0282851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0342s] [ 77%] 2024-08-06T21:15:52.0284094Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0295s] [ 77%] 2024-08-06T21:15:52.0285312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0467s] [ 77%] 2024-08-06T21:15:52.0286541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.2123s] [ 77%] 2024-08-06T21:15:52.0287775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0352s] [ 77%] 2024-08-06T21:15:52.0289058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0350s] [ 77%] 2024-08-06T21:15:52.0290318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0098s] [ 77%] 2024-08-06T21:15:52.0291535Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0098s] [ 77%] 2024-08-06T21:15:52.0292767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0076s] [ 77%] 2024-08-06T21:15:52.0293993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 78%] 2024-08-06T21:15:52.0295228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0125s] [ 78%] 2024-08-06T21:15:52.0296443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0125s] [ 78%] 2024-08-06T21:15:52.0297736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0090s] [ 78%] 2024-08-06T21:15:52.0299012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0088s] [ 78%] 2024-08-06T21:15:52.0300232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0101s] [ 78%] 2024-08-06T21:15:52.0301440Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 78%] 2024-08-06T21:15:52.0302650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0079s] [ 78%] 2024-08-06T21:15:52.0303880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 78%] 2024-08-06T21:15:52.0305571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0127s] [ 78%] 2024-08-06T21:15:52.0306805Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0127s] [ 78%] 2024-08-06T21:15:52.0308149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0088s] [ 78%] 2024-08-06T21:15:52.0309459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0088s] [ 78%] 2024-08-06T21:15:52.0310674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0105s] [ 78%] 2024-08-06T21:15:52.0311917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0106s] [ 78%] 2024-08-06T21:15:52.0313138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0084s] [ 78%] 2024-08-06T21:15:52.0314355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 78%] 2024-08-06T21:15:52.0315590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0131s] [ 78%] 2024-08-06T21:15:52.0316877Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0132s] [ 78%] 2024-08-06T21:15:52.0318190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0092s] [ 78%] 2024-08-06T21:15:52.0319626Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0092s] [ 78%] 2024-08-06T21:15:52.0321227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0107s] [ 78%] 2024-08-06T21:15:52.0323598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0110s] [ 78%] 2024-08-06T21:15:52.0325896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0085s] [ 78%] 2024-08-06T21:15:52.0328192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0085s] [ 78%] 2024-08-06T21:15:52.0330503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0134s] [ 78%] 2024-08-06T21:15:52.0332879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0131s] [ 78%] 2024-08-06T21:15:52.0335248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0089s] [ 78%] 2024-08-06T21:15:52.0337539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0089s] [ 78%] 2024-08-06T21:15:52.0339861Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0121s] [ 78%] 2024-08-06T21:15:52.0342174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0124s] [ 78%] 2024-08-06T21:15:52.0344483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0090s] [ 78%] 2024-08-06T21:15:52.0346803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0090s] [ 78%] 2024-08-06T21:15:52.0349146Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0148s] [ 78%] 2024-08-06T21:15:52.0351521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0149s] [ 78%] 2024-08-06T21:15:52.0353859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0100s] [ 78%] 2024-08-06T21:15:52.0356183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0100s] [ 78%] 2024-08-06T21:15:52.0358496Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0124s] [ 78%] 2024-08-06T21:15:52.0360800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0127s] [ 78%] 2024-08-06T21:15:52.0363339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0092s] [ 78%] 2024-08-06T21:15:52.0365653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0092s] [ 78%] 2024-08-06T21:15:52.0368089Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0150s] [ 78%] 2024-08-06T21:15:52.0370463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0151s] [ 78%] 2024-08-06T21:15:52.0372781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0103s] [ 78%] 2024-08-06T21:15:52.0375082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.1815s] [ 78%] 2024-08-06T21:15:52.0377398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0094s] [ 78%] 2024-08-06T21:15:52.0379708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0093s] [ 78%] 2024-08-06T21:15:52.0382015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0072s] [ 78%] 2024-08-06T21:15:52.0384374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 78%] 2024-08-06T21:15:52.0386732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0119s] [ 78%] 2024-08-06T21:15:52.0389046Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0119s] [ 78%] 2024-08-06T21:15:52.0391429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0082s] [ 78%] 2024-08-06T21:15:52.0393773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 78%] 2024-08-06T21:15:52.0396092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0096s] [ 78%] 2024-08-06T21:15:52.0398369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0096s] [ 78%] 2024-08-06T21:15:52.0400690Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0074s] [ 78%] 2024-08-06T21:15:52.0403043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 78%] 2024-08-06T21:15:52.0405859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0123s] [ 78%] 2024-08-06T21:15:52.0408174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0123s] [ 78%] 2024-08-06T21:15:52.0410495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0085s] [ 78%] 2024-08-06T21:15:52.0412792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 78%] 2024-08-06T21:15:52.0415092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0079s] [ 78%] 2024-08-06T21:15:52.0417399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 78%] 2024-08-06T21:15:52.0419800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0066s] [ 78%] 2024-08-06T21:15:52.0422207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 78%] 2024-08-06T21:15:52.0424504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0103s] [ 78%] 2024-08-06T21:15:52.0426840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 78%] 2024-08-06T21:15:52.0429181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0075s] [ 78%] 2024-08-06T21:15:52.0431505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 78%] 2024-08-06T21:15:52.0433817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0081s] [ 78%] 2024-08-06T21:15:52.0436451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0082s] [ 78%] 2024-08-06T21:15:52.0438841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0068s] [ 78%] 2024-08-06T21:15:52.0441227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 78%] 2024-08-06T21:15:52.0443538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0106s] [ 78%] 2024-08-06T21:15:52.0445841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 78%] 2024-08-06T21:15:52.0448140Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0077s] [ 78%] 2024-08-06T21:15:52.0450452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 78%] 2024-08-06T21:15:52.0452751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0081s] [ 79%] 2024-08-06T21:15:52.0455109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0081s] [ 79%] 2024-08-06T21:15:52.0457465Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0065s] [ 79%] 2024-08-06T21:15:52.0459792Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 79%] 2024-08-06T21:15:52.0462084Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0105s] [ 79%] 2024-08-06T21:15:52.0464401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 79%] 2024-08-06T21:15:52.0466714Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0072s] [ 79%] 2024-08-06T21:15:52.0469032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 79%] 2024-08-06T21:15:52.0471339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0083s] [ 79%] 2024-08-06T21:15:52.0473663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0083s] [ 79%] 2024-08-06T21:15:52.0476003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0068s] [ 79%] 2024-08-06T21:15:52.0478305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 79%] 2024-08-06T21:15:52.0480606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0107s] [ 79%] 2024-08-06T21:15:52.0482915Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0107s] [ 79%] 2024-08-06T21:15:52.0485217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0076s] [ 79%] 2024-08-06T21:15:52.0487525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 79%] 2024-08-06T21:15:52.0489867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0102s] [ 79%] 2024-08-06T21:15:52.0492239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 79%] 2024-08-06T21:15:52.0494541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0076s] [ 79%] 2024-08-06T21:15:52.0496831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 79%] 2024-08-06T21:15:52.0499148Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0122s] [ 79%] 2024-08-06T21:15:52.0501480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0122s] [ 79%] 2024-08-06T21:15:52.0503789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0082s] [ 79%] 2024-08-06T21:15:52.0506599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0081s] [ 79%] 2024-08-06T21:15:52.0509002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0103s] [ 79%] 2024-08-06T21:15:52.0511369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0112s] [ 79%] 2024-08-06T21:15:52.0513660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0118s] [ 79%] 2024-08-06T21:15:52.0515959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0117s] [ 79%] 2024-08-06T21:15:52.0518289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0162s] [ 79%] 2024-08-06T21:15:52.0520598Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0209s] [ 79%] 2024-08-06T21:15:52.0522888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0209s] [ 79%] 2024-08-06T21:15:52.0525272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 79%] 2024-08-06T21:15:52.0527643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0070s] [ 79%] 2024-08-06T21:15:52.0529947Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0070s] [ 79%] 2024-08-06T21:15:52.0532246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 79%] 2024-08-06T21:15:52.0534542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 79%] 2024-08-06T21:15:52.0536869Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0094s] [ 79%] 2024-08-06T21:15:52.0539182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0094s] [ 79%] 2024-08-06T21:15:52.0541491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 79%] 2024-08-06T21:15:52.0543884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 79%] 2024-08-06T21:15:52.0546223Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0073s] [ 79%] 2024-08-06T21:15:52.0548520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 79%] 2024-08-06T21:15:52.0550833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 79%] 2024-08-06T21:15:52.0553138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 79%] 2024-08-06T21:15:52.0555448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0096s] [ 79%] 2024-08-06T21:15:52.0557749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0096s] [ 79%] 2024-08-06T21:15:52.0560082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 79%] 2024-08-06T21:15:52.0562453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_256_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 79%] 2024-08-06T21:15:52.0564759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 79%] 2024-08-06T21:15:52.0567064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0050s] [ 79%] 2024-08-06T21:15:52.0569387Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 79%] 2024-08-06T21:15:52.0571687Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 79%] 2024-08-06T21:15:52.0574000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 79%] 2024-08-06T21:15:52.0576341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 79%] 2024-08-06T21:15:52.0578698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 79%] 2024-08-06T21:15:52.0581087Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 79%] 2024-08-06T21:15:52.0583370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 79%] 2024-08-06T21:15:52.0585664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 79%] 2024-08-06T21:15:52.0587961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 79%] 2024-08-06T21:15:52.0590256Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 79%] 2024-08-06T21:15:52.0592549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 79%] 2024-08-06T21:15:52.0594878Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 79%] 2024-08-06T21:15:52.0597228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 79%] 2024-08-06T21:15:52.0599553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 79%] 2024-08-06T21:15:52.0601874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 79%] 2024-08-06T21:15:52.0604593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 79%] 2024-08-06T21:15:52.0606925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 79%] 2024-08-06T21:15:52.0609242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 79%] 2024-08-06T21:15:52.0611552Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 79%] 2024-08-06T21:15:52.0613945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 79%] 2024-08-06T21:15:52.0616331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 79%] 2024-08-06T21:15:52.0618662Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 79%] 2024-08-06T21:15:52.0620945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 79%] 2024-08-06T21:15:52.0623236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 79%] 2024-08-06T21:15:52.0625524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 79%] 2024-08-06T21:15:52.0627815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 79%] 2024-08-06T21:15:52.0630164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 79%] 2024-08-06T21:15:52.0632516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 80%] 2024-08-06T21:15:52.0634845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 80%] 2024-08-06T21:15:52.0637168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 80%] 2024-08-06T21:15:52.0639480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 80%] 2024-08-06T21:15:52.0641794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 80%] 2024-08-06T21:15:52.0644080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 80%] 2024-08-06T21:15:52.0646388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 80%] 2024-08-06T21:15:52.0648739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 80%] 2024-08-06T21:15:52.0651111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 80%] 2024-08-06T21:15:52.0653514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 80%] 2024-08-06T21:15:52.0655831Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 80%] 2024-08-06T21:15:52.0658136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 80%] 2024-08-06T21:15:52.0660453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 80%] 2024-08-06T21:15:52.0662739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 80%] 2024-08-06T21:15:52.0665081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 80%] 2024-08-06T21:15:52.0667436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 80%] 2024-08-06T21:15:52.0669716Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 80%] 2024-08-06T21:15:52.0672009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 80%] 2024-08-06T21:15:52.0674338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 80%] 2024-08-06T21:15:52.0676645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0049s] [ 80%] 2024-08-06T21:15:52.0678941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0049s] [ 80%] 2024-08-06T21:15:52.0681217Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 80%] 2024-08-06T21:15:52.0683559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 80%] 2024-08-06T21:15:52.0685941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0057s] [ 80%] 2024-08-06T21:15:52.0688245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 80%] 2024-08-06T21:15:52.0690655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 80%] 2024-08-06T21:15:52.0692993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 80%] 2024-08-06T21:15:52.0695295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 80%] 2024-08-06T21:15:52.0697585Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 80%] 2024-08-06T21:15:52.0699912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 80%] 2024-08-06T21:15:52.0702253Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 80%] 2024-08-06T21:15:52.0704975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 80%] 2024-08-06T21:15:52.0707300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 80%] 2024-08-06T21:15:52.0709592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 80%] 2024-08-06T21:15:52.0711913Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 80%] 2024-08-06T21:15:52.0714214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0062s] [ 80%] 2024-08-06T21:15:52.0716520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 80%] 2024-08-06T21:15:52.0718912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 80%] 2024-08-06T21:15:52.0721293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 80%] 2024-08-06T21:15:52.0723600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 80%] 2024-08-06T21:15:52.0725922Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 80%] 2024-08-06T21:15:52.0728236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 80%] 2024-08-06T21:15:52.0730550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 80%] 2024-08-06T21:15:52.0732856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 80%] 2024-08-06T21:15:52.0735152Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 80%] 2024-08-06T21:15:52.0737506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 80%] 2024-08-06T21:15:52.0739872Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 80%] 2024-08-06T21:15:52.0742169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 80%] 2024-08-06T21:15:52.0744480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 80%] 2024-08-06T21:15:52.0746789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 80%] 2024-08-06T21:15:52.0749082Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 80%] 2024-08-06T21:15:52.0751410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0069s] [ 80%] 2024-08-06T21:15:52.0753767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 80%] 2024-08-06T21:15:52.0756130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 80%] 2024-08-06T21:15:52.0758438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 80%] 2024-08-06T21:15:52.0760726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0075s] [ 80%] 2024-08-06T21:15:52.0763050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 80%] 2024-08-06T21:15:52.0765368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 80%] 2024-08-06T21:15:52.0767684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 80%] 2024-08-06T21:15:52.0769994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 80%] 2024-08-06T21:15:52.0772316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 80%] 2024-08-06T21:15:52.0774660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 80%] 2024-08-06T21:15:52.0776976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 80%] 2024-08-06T21:15:52.0779278Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 80%] 2024-08-06T21:15:52.0781584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 80%] 2024-08-06T21:15:52.0783879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 80%] 2024-08-06T21:15:52.0786197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 80%] 2024-08-06T21:15:52.0788542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0086s] [ 80%] 2024-08-06T21:15:52.0790907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0086s] [ 80%] 2024-08-06T21:15:52.0793215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 80%] 2024-08-06T21:15:52.0795516Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 80%] 2024-08-06T21:15:52.0797807Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0092s] [ 80%] 2024-08-06T21:15:52.0800128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0091s] [ 80%] 2024-08-06T21:15:52.0802439Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 80%] 2024-08-06T21:15:52.0805196Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 80%] 2024-08-06T21:15:52.0807623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 80%] 2024-08-06T21:15:52.0810010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 80%] 2024-08-06T21:15:52.0812304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 80%] 2024-08-06T21:15:52.0814603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 81%] 2024-08-06T21:15:52.0816911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0071s] [ 81%] 2024-08-06T21:15:52.0819221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 81%] 2024-08-06T21:15:52.0821582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0067s] [ 81%] 2024-08-06T21:15:52.0823888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 81%] 2024-08-06T21:15:52.0826273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0059s] [ 81%] 2024-08-06T21:15:52.0828643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 81%] 2024-08-06T21:15:52.0830939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 81%] 2024-08-06T21:15:52.0833225Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 81%] 2024-08-06T21:15:52.0835559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 81%] 2024-08-06T21:15:52.0837865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 81%] 2024-08-06T21:15:52.0840168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 81%] 2024-08-06T21:15:52.0842532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 81%] 2024-08-06T21:15:52.0844894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 81%] 2024-08-06T21:15:52.0847162Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 81%] 2024-08-06T21:15:52.0849451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 81%] 2024-08-06T21:15:52.0851751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 81%] 2024-08-06T21:15:52.0854059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 81%] 2024-08-06T21:15:52.0856355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 81%] 2024-08-06T21:15:52.0858634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 81%] 2024-08-06T21:15:52.0860980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 81%] 2024-08-06T21:15:52.0863338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0049s] [ 81%] 2024-08-06T21:15:52.0865625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0049s] [ 81%] 2024-08-06T21:15:52.0867939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 81%] 2024-08-06T21:15:52.0870218Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0047s] [ 81%] 2024-08-06T21:15:52.0872512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0057s] [ 81%] 2024-08-06T21:15:52.0874806Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 81%] 2024-08-06T21:15:52.0877212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0055s] [ 81%] 2024-08-06T21:15:52.0879562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 81%] 2024-08-06T21:15:52.0881844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 81%] 2024-08-06T21:15:52.0884125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 81%] 2024-08-06T21:15:52.0886411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 81%] 2024-08-06T21:15:52.0888696Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 81%] 2024-08-06T21:15:52.0890979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 81%] 2024-08-06T21:15:52.0893287Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 81%] 2024-08-06T21:15:52.0895617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 81%] 2024-08-06T21:15:52.0897959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 81%] 2024-08-06T21:15:52.0900268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0049s] [ 81%] 2024-08-06T21:15:52.0902627Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0049s] [ 81%] 2024-08-06T21:15:52.0905402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 81%] 2024-08-06T21:15:52.0907698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 81%] 2024-08-06T21:15:52.0909996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 81%] 2024-08-06T21:15:52.0912393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 81%] 2024-08-06T21:15:52.0914809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 81%] 2024-08-06T21:15:52.0917107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 81%] 2024-08-06T21:15:52.0919413Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 81%] 2024-08-06T21:15:52.0921703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 81%] 2024-08-06T21:15:52.0923991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 81%] 2024-08-06T21:15:52.0926268Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 81%] 2024-08-06T21:15:52.0928547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 81%] 2024-08-06T21:15:52.0930879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 81%] 2024-08-06T21:15:52.0933239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 81%] 2024-08-06T21:15:52.0935521Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 81%] 2024-08-06T21:15:52.0938581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0049s] [ 81%] 2024-08-06T21:15:52.0942569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0050s] [ 81%] 2024-08-06T21:15:52.0946819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 81%] 2024-08-06T21:15:52.0951518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0047s] [ 81%] 2024-08-06T21:15:52.0956480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0057s] [ 81%] 2024-08-06T21:15:52.0961478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 81%] 2024-08-06T21:15:52.0966396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0055s] [ 81%] 2024-08-06T21:15:52.0971239Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 81%] 2024-08-06T21:15:52.0976125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 81%] 2024-08-06T21:15:52.0980961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 81%] 2024-08-06T21:15:52.0985896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 81%] 2024-08-06T21:15:52.0990817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 81%] 2024-08-06T21:15:52.0995776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 81%] 2024-08-06T21:15:52.1000673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 81%] 2024-08-06T21:15:52.1005980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 81%] 2024-08-06T21:15:52.1010933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 81%] 2024-08-06T21:15:52.1015821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0049s] [ 81%] 2024-08-06T21:15:52.1020684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0049s] [ 81%] 2024-08-06T21:15:52.1025506Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0047s] [ 81%] 2024-08-06T21:15:52.1030518Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0047s] [ 81%] 2024-08-06T21:15:52.1035471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0057s] [ 81%] 2024-08-06T21:15:52.1040320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 81%] 2024-08-06T21:15:52.1045158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0055s] [ 81%] 2024-08-06T21:15:52.1049996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 81%] 2024-08-06T21:15:52.1054817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 82%] 2024-08-06T21:15:52.1059675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 82%] 2024-08-06T21:15:52.1064471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 82%] 2024-08-06T21:15:52.1069525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 82%] 2024-08-06T21:15:52.1074403Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 82%] 2024-08-06T21:15:52.1079220Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 82%] 2024-08-06T21:15:52.1084045Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 82%] 2024-08-06T21:15:52.1088833Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 82%] 2024-08-06T21:15:52.1093788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0093s] [ 82%] 2024-08-06T21:15:52.1098655Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0094s] [ 82%] 2024-08-06T21:15:52.1103526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 82%] 2024-08-06T21:15:52.1108881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 82%] 2024-08-06T21:15:52.1113933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0105s] [ 82%] 2024-08-06T21:15:52.1118824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 82%] 2024-08-06T21:15:52.1123713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 82%] 2024-08-06T21:15:52.1128614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 82%] 2024-08-06T21:15:52.1133602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 82%] 2024-08-06T21:15:52.1138453Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 82%] 2024-08-06T21:15:52.1143491Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 82%] 2024-08-06T21:15:52.1148422Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 82%] 2024-08-06T21:15:52.1153273Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 82%] 2024-08-06T21:15:52.1158155Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 82%] 2024-08-06T21:15:52.1163049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 82%] 2024-08-06T21:15:52.1167930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0064s] [ 82%] 2024-08-06T21:15:52.1172823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0109s] [ 82%] 2024-08-06T21:15:52.1177713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0109s] [ 82%] 2024-08-06T21:15:52.1182740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0062s] [ 82%] 2024-08-06T21:15:52.1187707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 82%] 2024-08-06T21:15:52.1192612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0120s] [ 82%] 2024-08-06T21:15:52.1197512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0121s] [ 82%] 2024-08-06T21:15:52.1202427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 82%] 2024-08-06T21:15:52.1207793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 82%] 2024-08-06T21:15:52.1212669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0069s] [ 82%] 2024-08-06T21:15:52.1217348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 82%] 2024-08-06T21:15:52.1221467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0065s] [ 82%] 2024-08-06T21:15:52.1225773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 82%] 2024-08-06T21:15:52.1230312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0076s] [ 82%] 2024-08-06T21:15:52.1234749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 82%] 2024-08-06T21:15:52.1239341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0073s] [ 82%] 2024-08-06T21:15:52.1242726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0073s] [ 82%] 2024-08-06T21:15:52.1245069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0137s] [ 82%] 2024-08-06T21:15:52.1247547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0140s] [ 82%] 2024-08-06T21:15:52.1249950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0080s] [ 82%] 2024-08-06T21:15:52.1252258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 82%] 2024-08-06T21:15:52.1254576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0150s] [ 82%] 2024-08-06T21:15:52.1256903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0151s] [ 82%] 2024-08-06T21:15:52.1259205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0089s] [ 82%] 2024-08-06T21:15:52.1261517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0089s] [ 82%] 2024-08-06T21:15:52.1263825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0085s] [ 82%] 2024-08-06T21:15:52.1266190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 82%] 2024-08-06T21:15:52.1268530Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0076s] [ 82%] 2024-08-06T21:15:52.1270821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0076s] [ 82%] 2024-08-06T21:15:52.1273117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0092s] [ 82%] 2024-08-06T21:15:52.1275423Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0092s] [ 82%] 2024-08-06T21:15:52.1277717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0086s] [ 82%] 2024-08-06T21:15:52.1280011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 82%] 2024-08-06T21:15:52.1282340Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0086s] [ 82%] 2024-08-06T21:15:52.1284712Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0086s] [ 82%] 2024-08-06T21:15:52.1287003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 82%] 2024-08-06T21:15:52.1289298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 82%] 2024-08-06T21:15:52.1291673Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0096s] [ 82%] 2024-08-06T21:15:52.1293985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0097s] [ 82%] 2024-08-06T21:15:52.1296280Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 82%] 2024-08-06T21:15:52.1298587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 82%] 2024-08-06T21:15:52.1300929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 82%] 2024-08-06T21:15:52.1303332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 82%] 2024-08-06T21:15:52.1306142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 82%] 2024-08-06T21:15:52.1308427Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 82%] 2024-08-06T21:15:52.1310725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 82%] 2024-08-06T21:15:52.1313013Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 82%] 2024-08-06T21:15:52.1315304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 82%] 2024-08-06T21:15:52.1317679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 82%] 2024-08-06T21:15:52.1320029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 82%] 2024-08-06T21:15:52.1322376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0049s] [ 82%] 2024-08-06T21:15:52.1324670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 82%] 2024-08-06T21:15:52.1326970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 82%] 2024-08-06T21:15:52.1329306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0057s] [ 82%] 2024-08-06T21:15:52.1331596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 82%] 2024-08-06T21:15:52.1333898Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 83%] 2024-08-06T21:15:52.1336271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 83%] 2024-08-06T21:15:52.1339371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 83%] 2024-08-06T21:15:52.1341669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 83%] 2024-08-06T21:15:52.1343959Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 83%] 2024-08-06T21:15:52.1346229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 83%] 2024-08-06T21:15:52.1348522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 83%] 2024-08-06T21:15:52.1350811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 83%] 2024-08-06T21:15:52.1353150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 83%] 2024-08-06T21:15:52.1355522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 83%] 2024-08-06T21:15:52.1357796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0049s] [ 83%] 2024-08-06T21:15:52.1360107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0049s] [ 83%] 2024-08-06T21:15:52.1362429Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 83%] 2024-08-06T21:15:52.1364728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 83%] 2024-08-06T21:15:52.1367023Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0057s] [ 83%] 2024-08-06T21:15:52.1369303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 83%] 2024-08-06T21:15:52.1371656Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 83%] 2024-08-06T21:15:52.1374029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 83%] 2024-08-06T21:15:52.1376311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 83%] 2024-08-06T21:15:52.1378607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 83%] 2024-08-06T21:15:52.1380894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 83%] 2024-08-06T21:15:52.1383172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 83%] 2024-08-06T21:15:52.1385455Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 83%] 2024-08-06T21:15:52.1387808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 83%] 2024-08-06T21:15:52.1390145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 83%] 2024-08-06T21:15:52.1392404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 83%] 2024-08-06T21:15:52.1394684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 83%] 2024-08-06T21:15:52.1396985Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0050s] [ 83%] 2024-08-06T21:15:52.1399286Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 83%] 2024-08-06T21:15:52.1401575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0049s] [ 83%] 2024-08-06T21:15:52.1404620Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 83%] 2024-08-06T21:15:52.1407074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 83%] 2024-08-06T21:15:52.1409466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 83%] 2024-08-06T21:15:52.1411770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 83%] 2024-08-06T21:15:52.1414081Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 83%] 2024-08-06T21:15:52.1416370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 83%] 2024-08-06T21:15:52.1418634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 83%] 2024-08-06T21:15:52.1420914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 83%] 2024-08-06T21:15:52.1423258Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 83%] 2024-08-06T21:15:52.1425612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 83%] 2024-08-06T21:15:52.1427901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 83%] 2024-08-06T21:15:52.1430171Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 83%] 2024-08-06T21:15:52.1432470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 83%] 2024-08-06T21:15:52.1434768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 83%] 2024-08-06T21:15:52.1437067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 83%] 2024-08-06T21:15:52.1439376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 83%] 2024-08-06T21:15:52.1441700Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 83%] 2024-08-06T21:15:52.1444433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 83%] 2024-08-06T21:15:52.1447000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 83%] 2024-08-06T21:15:52.1449300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 83%] 2024-08-06T21:15:52.1451590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 83%] 2024-08-06T21:15:52.1453856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 83%] 2024-08-06T21:15:52.1456142Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 83%] 2024-08-06T21:15:52.1458490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 83%] 2024-08-06T21:15:52.1460829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 83%] 2024-08-06T21:15:52.1463110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 83%] 2024-08-06T21:15:52.1465404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 83%] 2024-08-06T21:15:52.1467669Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0100s] [ 83%] 2024-08-06T21:15:52.1469970Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0091s] [ 83%] 2024-08-06T21:15:52.1472276Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0090s] [ 83%] 2024-08-06T21:15:52.1474584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0088s] [ 83%] 2024-08-06T21:15:52.1476923Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0088s] [ 83%] 2024-08-06T21:15:52.1479248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0106s] [ 83%] 2024-08-06T21:15:52.1481542Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0106s] [ 83%] 2024-08-06T21:15:52.1483837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0103s] [ 83%] 2024-08-06T21:15:52.1486132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0104s] [ 83%] 2024-08-06T21:15:52.1488421Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0072s] [ 83%] 2024-08-06T21:15:52.1490702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 83%] 2024-08-06T21:15:52.1492998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 83%] 2024-08-06T21:15:52.1495324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 83%] 2024-08-06T21:15:52.1497667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 83%] 2024-08-06T21:15:52.1499950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 83%] 2024-08-06T21:15:52.1502222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 83%] 2024-08-06T21:15:52.1504953Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 83%] 2024-08-06T21:15:52.1507281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0049s] [ 83%] 2024-08-06T21:15:52.1509575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0049s] [ 83%] 2024-08-06T21:15:52.1511994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 83%] 2024-08-06T21:15:52.1514349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0047s] [ 84%] 2024-08-06T21:15:52.1516657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0057s] [ 84%] 2024-08-06T21:15:52.1518951Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 84%] 2024-08-06T21:15:52.1521252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 84%] 2024-08-06T21:15:52.1523576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 84%] 2024-08-06T21:15:52.1525860Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 84%] 2024-08-06T21:15:52.1528129Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 84%] 2024-08-06T21:15:52.1530489Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 84%] 2024-08-06T21:15:52.1532848Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 84%] 2024-08-06T21:15:52.1535135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 84%] 2024-08-06T21:15:52.1537412Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 84%] 2024-08-06T21:15:52.1539686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 84%] 2024-08-06T21:15:52.1541995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 84%] 2024-08-06T21:15:52.1544283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0049s] [ 84%] 2024-08-06T21:15:52.1546582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0049s] [ 84%] 2024-08-06T21:15:52.1548956Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 84%] 2024-08-06T21:15:52.1551277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 84%] 2024-08-06T21:15:52.1553569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0057s] [ 84%] 2024-08-06T21:15:52.1555867Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 84%] 2024-08-06T21:15:52.1558168Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0055s] [ 84%] 2024-08-06T21:15:52.1560466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 84%] 2024-08-06T21:15:52.1562729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 84%] 2024-08-06T21:15:52.1565051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 84%] 2024-08-06T21:15:52.1567388Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 84%] 2024-08-06T21:15:52.1569665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 84%] 2024-08-06T21:15:52.1571954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 84%] 2024-08-06T21:15:52.1574248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 84%] 2024-08-06T21:15:52.1576534Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 84%] 2024-08-06T21:15:52.1578819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 84%] 2024-08-06T21:15:52.1581103Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0049s] [ 84%] 2024-08-06T21:15:52.1583435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0049s] [ 84%] 2024-08-06T21:15:52.1585766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 84%] 2024-08-06T21:15:52.1588029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 84%] 2024-08-06T21:15:52.1590365Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 84%] 2024-08-06T21:15:52.1592670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 84%] 2024-08-06T21:15:52.1595014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 84%] 2024-08-06T21:15:52.1597301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 84%] 2024-08-06T21:15:52.1599630Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 84%] 2024-08-06T21:15:52.1601958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 84%] 2024-08-06T21:15:52.1604619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 84%] 2024-08-06T21:15:52.1606965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 84%] 2024-08-06T21:15:52.1609251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 84%] 2024-08-06T21:15:52.1611514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 84%] 2024-08-06T21:15:52.1613790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 84%] 2024-08-06T21:15:52.1616069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_4_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 84%] 2024-08-06T21:15:52.1618470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0268s] [ 84%] 2024-08-06T21:15:52.1620874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0277s] [ 84%] 2024-08-06T21:15:52.1623183Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0169s] [ 84%] 2024-08-06T21:15:52.1625550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0169s] [ 84%] 2024-08-06T21:15:52.1627888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0379s] [ 84%] 2024-08-06T21:15:52.1630229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0383s] [ 84%] 2024-08-06T21:15:52.1632563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0204s] [ 84%] 2024-08-06T21:15:52.1635005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0204s] [ 84%] 2024-08-06T21:15:52.1637404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0268s] [ 84%] 2024-08-06T21:15:52.1639720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0276s] [ 84%] 2024-08-06T21:15:52.1642033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0174s] [ 84%] 2024-08-06T21:15:52.1644373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0174s] [ 84%] 2024-08-06T21:15:52.1646702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.2045s] [ 84%] 2024-08-06T21:15:52.1649012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0383s] [ 84%] 2024-08-06T21:15:52.1650252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0218s] [ 84%] 2024-08-06T21:15:52.1651526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0218s] [ 84%] 2024-08-06T21:15:52.1652823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0304s] [ 84%] 2024-08-06T21:15:52.1654050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0315s] [ 84%] 2024-08-06T21:15:52.1655289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0198s] [ 84%] 2024-08-06T21:15:52.1656541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0198s] [ 84%] 2024-08-06T21:15:52.1657774Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0409s] [ 84%] 2024-08-06T21:15:52.1659020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0412s] [ 84%] 2024-08-06T21:15:52.1660293Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0231s] [ 84%] 2024-08-06T21:15:52.1661596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0231s] [ 84%] 2024-08-06T21:15:52.1662814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0297s] [ 84%] 2024-08-06T21:15:52.1664051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0308s] [ 84%] 2024-08-06T21:15:52.1665271Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0204s] [ 84%] 2024-08-06T21:15:52.1666529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0205s] [ 84%] 2024-08-06T21:15:52.1667747Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0404s] [ 84%] 2024-08-06T21:15:52.1668977Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0407s] [ 84%] 2024-08-06T21:15:52.1670254Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0236s] [ 84%] 2024-08-06T21:15:52.1671533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0241s] [ 84%] 2024-08-06T21:15:52.1672772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.1997s] [ 85%] 2024-08-06T21:15:52.1674003Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0369s] [ 85%] 2024-08-06T21:15:52.1675250Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0236s] [ 85%] 2024-08-06T21:15:52.1676490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0237s] [ 85%] 2024-08-06T21:15:52.1677732Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0458s] [ 85%] 2024-08-06T21:15:52.1679010Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0459s] [ 85%] 2024-08-06T21:15:52.1680308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0274s] [ 85%] 2024-08-06T21:15:52.1681537Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0273s] [ 85%] 2024-08-06T21:15:52.1682756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0347s] [ 85%] 2024-08-06T21:15:52.1684002Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0365s] [ 85%] 2024-08-06T21:15:52.1685228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0248s] [ 85%] 2024-08-06T21:15:52.1686475Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0247s] [ 85%] 2024-08-06T21:15:52.1687695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0459s] [ 85%] 2024-08-06T21:15:52.1688980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0459s] [ 85%] 2024-08-06T21:15:52.1690248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0280s] [ 85%] 2024-08-06T21:15:52.1691493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.1882s] [ 85%] 2024-08-06T21:15:52.1692719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0255s] [ 85%] 2024-08-06T21:15:52.1693952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0262s] [ 85%] 2024-08-06T21:15:52.1695188Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0164s] [ 85%] 2024-08-06T21:15:52.1696424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0164s] [ 85%] 2024-08-06T21:15:52.1697701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0368s] [ 85%] 2024-08-06T21:15:52.1698978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0371s] [ 85%] 2024-08-06T21:15:52.1700222Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0199s] [ 85%] 2024-08-06T21:15:52.1701449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0198s] [ 85%] 2024-08-06T21:15:52.1702683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0257s] [ 85%] 2024-08-06T21:15:52.1704192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0264s] [ 85%] 2024-08-06T21:15:52.1705567Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0171s] [ 85%] 2024-08-06T21:15:52.1706818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0171s] [ 85%] 2024-08-06T21:15:52.1708119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0364s] [ 85%] 2024-08-06T21:15:52.1709437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0371s] [ 85%] 2024-08-06T21:15:52.1710653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0211s] [ 85%] 2024-08-06T21:15:52.1711894Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0210s] [ 85%] 2024-08-06T21:15:52.1713122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0509s] [ 85%] 2024-08-06T21:15:52.1714369Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0537s] [ 85%] 2024-08-06T21:15:52.1715589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0308s] [ 85%] 2024-08-06T21:15:52.1716895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0304s] [ 85%] 2024-08-06T21:15:52.1718206Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.2412s] [ 85%] 2024-08-06T21:15:52.1719436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0739s] [ 85%] 2024-08-06T21:15:52.1720676Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0374s] [ 85%] 2024-08-06T21:15:52.1721912Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0373s] [ 85%] 2024-08-06T21:15:52.1723150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0478s] [ 85%] 2024-08-06T21:15:52.1724366Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0503s] [ 85%] 2024-08-06T21:15:52.1725597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0327s] [ 85%] 2024-08-06T21:15:52.1726876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0316s] [ 85%] 2024-08-06T21:15:52.1728154Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0674s] [ 85%] 2024-08-06T21:15:52.1729399Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0691s] [ 85%] 2024-08-06T21:15:52.1730624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.1958s] [ 85%] 2024-08-06T21:15:52.1731871Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0424s] [ 85%] 2024-08-06T21:15:52.1733093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0582s] [ 85%] 2024-08-06T21:15:52.1734333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0615s] [ 85%] 2024-08-06T21:15:52.1735595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0375s] [ 85%] 2024-08-06T21:15:52.1736903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0378s] [ 85%] 2024-08-06T21:15:52.1738132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0804s] [ 85%] 2024-08-06T21:15:52.1739363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0818s] [ 85%] 2024-08-06T21:15:52.1740611Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0453s] [ 85%] 2024-08-06T21:15:52.1741847Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.1993s] [ 85%] 2024-08-06T21:15:52.1743080Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0569s] [ 85%] 2024-08-06T21:15:52.1744301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0603s] [ 85%] 2024-08-06T21:15:52.1745577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0396s] [ 85%] 2024-08-06T21:15:52.1746855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0396s] [ 85%] 2024-08-06T21:15:52.1748092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0749s] [ 85%] 2024-08-06T21:15:52.1749326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0785s] [ 85%] 2024-08-06T21:15:52.1750551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0472s] [ 85%] 2024-08-06T21:15:52.1751798Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0468s] [ 85%] 2024-08-06T21:15:52.1753016Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.2218s] [ 85%] 2024-08-06T21:15:52.1754308Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0728s] [ 85%] 2024-08-06T21:15:52.1755583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0449s] [ 85%] 2024-08-06T21:15:52.1756837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0452s] [ 85%] 2024-08-06T21:15:52.1758064Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0925s] [ 85%] 2024-08-06T21:15:52.1759316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0933s] [ 85%] 2024-08-06T21:15:52.1760548Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0574s] [ 85%] 2024-08-06T21:15:52.1761780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0563s] [ 85%] 2024-08-06T21:15:52.1763014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.2232s] [ 85%] 2024-08-06T21:15:52.1764283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0679s] [ 85%] 2024-08-06T21:15:52.1765566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0450s] [ 85%] 2024-08-06T21:15:52.1766795Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0451s] [ 85%] 2024-08-06T21:15:52.1768036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0850s] [ 85%] 2024-08-06T21:15:52.1769267Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0818s] [ 85%] 2024-08-06T21:15:52.1770505Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0506s] [ 86%] 2024-08-06T21:15:52.1771728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0531s] [ 86%] 2024-08-06T21:15:52.1772988Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.2123s] [ 86%] 2024-08-06T21:15:52.1774283Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0531s] [ 86%] 2024-08-06T21:15:52.1775499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0327s] [ 86%] 2024-08-06T21:15:52.1776751Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0317s] [ 86%] 2024-08-06T21:15:52.1777979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0734s] [ 86%] 2024-08-06T21:15:52.1779231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0733s] [ 86%] 2024-08-06T21:15:52.1780449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0396s] [ 86%] 2024-08-06T21:15:52.1781697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0391s] [ 86%] 2024-08-06T21:15:52.1782948Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0501s] [ 86%] 2024-08-06T21:15:52.1784232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0498s] [ 86%] 2024-08-06T21:15:52.1785459Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0328s] [ 86%] 2024-08-06T21:15:52.1786691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0326s] [ 86%] 2024-08-06T21:15:52.1787935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.2331s] [ 86%] 2024-08-06T21:15:52.1789153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0697s] [ 86%] 2024-08-06T21:15:52.1790381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0401s] [ 86%] 2024-08-06T21:15:52.1791643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0401s] [ 86%] 2024-08-06T21:15:52.1792928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0127s] [ 86%] 2024-08-06T21:15:52.1794138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0121s] [ 86%] 2024-08-06T21:15:52.1795348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0087s] [ 86%] 2024-08-06T21:15:52.1796596Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0089s] [ 86%] 2024-08-06T21:15:52.1797820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0162s] [ 86%] 2024-08-06T21:15:52.1799058Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0163s] [ 86%] 2024-08-06T21:15:52.1800274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0097s] [ 86%] 2024-08-06T21:15:52.1801561Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0099s] [ 86%] 2024-08-06T21:15:52.1802812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0123s] [ 86%] 2024-08-06T21:15:52.1804433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0121s] [ 86%] 2024-08-06T21:15:52.1805708Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0092s] [ 86%] 2024-08-06T21:15:52.1806941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0093s] [ 86%] 2024-08-06T21:15:52.1808167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0166s] [ 86%] 2024-08-06T21:15:52.1809380Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0163s] [ 86%] 2024-08-06T21:15:52.1810677Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0101s] [ 86%] 2024-08-06T21:15:52.1811965Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0100s] [ 86%] 2024-08-06T21:15:52.1813195Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0141s] [ 86%] 2024-08-06T21:15:52.1814410Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0142s] [ 86%] 2024-08-06T21:15:52.1815644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0104s] [ 86%] 2024-08-06T21:15:52.1816885Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0105s] [ 86%] 2024-08-06T21:15:52.1818097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0184s] [ 86%] 2024-08-06T21:15:52.1819344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0181s] [ 86%] 2024-08-06T21:15:52.1820619Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0103s] [ 86%] 2024-08-06T21:15:52.1821921Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0104s] [ 86%] 2024-08-06T21:15:52.1823131Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0147s] [ 86%] 2024-08-06T21:15:52.1824360Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0152s] [ 86%] 2024-08-06T21:15:52.1825571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0105s] [ 86%] 2024-08-06T21:15:52.1826818Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0105s] [ 86%] 2024-08-06T21:15:52.1828027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0179s] [ 86%] 2024-08-06T21:15:52.1829282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0174s] [ 86%] 2024-08-06T21:15:52.1830564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0102s] [ 86%] 2024-08-06T21:15:52.1831775Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0106s] [ 86%] 2024-08-06T21:15:52.1832998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0183s] [ 86%] 2024-08-06T21:15:52.1834216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0183s] [ 86%] 2024-08-06T21:15:52.1835451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0130s] [ 86%] 2024-08-06T21:15:52.1836681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0127s] [ 86%] 2024-08-06T21:15:52.1837909Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0211s] [ 86%] 2024-08-06T21:15:52.1839174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0215s] [ 86%] 2024-08-06T21:15:52.1840461Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0134s] [ 86%] 2024-08-06T21:15:52.1841681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0134s] [ 86%] 2024-08-06T21:15:52.1842890Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0183s] [ 86%] 2024-08-06T21:15:52.1844122Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0179s] [ 86%] 2024-08-06T21:15:52.1845330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.1792s] [ 86%] 2024-08-06T21:15:52.1846573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0132s] [ 86%] 2024-08-06T21:15:52.1847825Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0211s] [ 86%] 2024-08-06T21:15:52.1849114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0220s] [ 86%] 2024-08-06T21:15:52.1850324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0137s] [ 86%] 2024-08-06T21:15:52.1851554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0142s] [ 86%] 2024-08-06T21:15:52.1852770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0112s] [ 86%] 2024-08-06T21:15:52.1853992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0116s] [ 86%] 2024-08-06T21:15:52.1855215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0077s] [ 86%] 2024-08-06T21:15:52.1856449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0087s] [ 86%] 2024-08-06T21:15:52.1857743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0156s] [ 86%] 2024-08-06T21:15:52.1859015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0157s] [ 86%] 2024-08-06T21:15:52.1860244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0096s] [ 86%] 2024-08-06T21:15:52.1861467Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0088s] [ 86%] 2024-08-06T21:15:52.1862689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0116s] [ 86%] 2024-08-06T21:15:52.1863906Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0109s] [ 86%] 2024-08-06T21:15:52.1865106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0078s] [ 86%] 2024-08-06T21:15:52.1866396Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 87%] 2024-08-06T21:15:52.1867661Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0147s] [ 87%] 2024-08-06T21:15:52.1868888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0147s] [ 87%] 2024-08-06T21:15:52.1870090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0087s] [ 87%] 2024-08-06T21:15:52.1871324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0087s] [ 87%] 2024-08-06T21:15:52.1872554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0986s] [ 87%] 2024-08-06T21:15:52.1873801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.1046s] [ 87%] 2024-08-06T21:15:52.1875020Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0561s] [ 87%] 2024-08-06T21:15:52.1876312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0561s] [ 87%] 2024-08-06T21:15:52.1877612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.3026s] [ 87%] 2024-08-06T21:15:52.1878849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.1463s] [ 87%] 2024-08-06T21:15:52.1880096Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0710s] [ 87%] 2024-08-06T21:15:52.1881332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0710s] [ 87%] 2024-08-06T21:15:52.1882562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0821s] [ 87%] 2024-08-06T21:15:52.1883778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.2357s] [ 87%] 2024-08-06T21:15:52.1885055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0618s] [ 87%] 2024-08-06T21:15:52.1886353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0627s] [ 87%] 2024-08-06T21:15:52.1887576Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.1217s] [ 87%] 2024-08-06T21:15:52.1888819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.1243s] [ 87%] 2024-08-06T21:15:52.1890091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.2169s] [ 87%] 2024-08-06T21:15:52.1891353Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0785s] [ 87%] 2024-08-06T21:15:52.1892571Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.1104s] [ 87%] 2024-08-06T21:15:52.1893808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.1176s] [ 87%] 2024-08-06T21:15:52.1895075Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0692s] [ 87%] 2024-08-06T21:15:52.1896401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.2265s] [ 87%] 2024-08-06T21:15:52.1897624Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.1567s] [ 87%] 2024-08-06T21:15:52.1898856Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.1592s] [ 87%] 2024-08-06T21:15:52.1900107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0861s] [ 87%] 2024-08-06T21:15:52.1901341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.2214s] [ 87%] 2024-08-06T21:15:52.1902577Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0978s] [ 87%] 2024-08-06T21:15:52.1903841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.1017s] [ 87%] 2024-08-06T21:15:52.1905597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0716s] [ 87%] 2024-08-06T21:15:52.1906850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0695s] [ 87%] 2024-08-06T21:15:52.1908088Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.2861s] [ 87%] 2024-08-06T21:15:52.1909318Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.1305s] [ 87%] 2024-08-06T21:15:52.1910540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0847s] [ 87%] 2024-08-06T21:15:52.1911778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0848s] [ 87%] 2024-08-06T21:15:52.1912996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.2729s] [ 87%] 2024-08-06T21:15:52.1914324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.1371s] [ 87%] 2024-08-06T21:15:52.1915616Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0812s] [ 87%] 2024-08-06T21:15:52.1916887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0811s] [ 87%] 2024-08-06T21:15:52.1918117Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.3103s] [ 87%] 2024-08-06T21:15:52.1919371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.1776s] [ 87%] 2024-08-06T21:15:52.1920599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.1000s] [ 87%] 2024-08-06T21:15:52.1921824Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0993s] [ 87%] 2024-08-06T21:15:52.1923114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.2512s] [ 87%] 2024-08-06T21:15:52.1924418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.1153s] [ 87%] 2024-08-06T21:15:52.1925646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0806s] [ 87%] 2024-08-06T21:15:52.1926884Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0832s] [ 87%] 2024-08-06T21:15:52.1928123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.2724s] [ 87%] 2024-08-06T21:15:52.1929354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.1453s] [ 87%] 2024-08-06T21:15:52.1930584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0996s] [ 87%] 2024-08-06T21:15:52.1931808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0988s] [ 87%] 2024-08-06T21:15:52.1933063Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.2488s] [ 87%] 2024-08-06T21:15:52.1934405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.1025s] [ 87%] 2024-08-06T21:15:52.1935623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0578s] [ 87%] 2024-08-06T21:15:52.1936891Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0568s] [ 87%] 2024-08-06T21:15:52.1938126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.1418s] [ 87%] 2024-08-06T21:15:52.1939371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.1410s] [ 87%] 2024-08-06T21:15:52.1940588Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.2254s] [ 87%] 2024-08-06T21:15:52.1941875Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0683s] [ 87%] 2024-08-06T21:15:52.1943143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0781s] [ 87%] 2024-08-06T21:15:52.1944363Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0801s] [ 87%] 2024-08-06T21:15:52.1945592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0548s] [ 87%] 2024-08-06T21:15:52.1946838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0547s] [ 87%] 2024-08-06T21:15:52.1948079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.2649s] [ 87%] 2024-08-06T21:15:52.1949302Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.1111s] [ 87%] 2024-08-06T21:15:52.1950533Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0693s] [ 87%] 2024-08-06T21:15:52.1951796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0697s] [ 87%] 2024-08-06T21:15:52.1953079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0156s] [ 87%] 2024-08-06T21:15:52.1954299Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0159s] [ 87%] 2024-08-06T21:15:52.1955512Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0109s] [ 87%] 2024-08-06T21:15:52.1956781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0109s] [ 87%] 2024-08-06T21:15:52.1958005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0212s] [ 87%] 2024-08-06T21:15:52.1959247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0213s] [ 87%] 2024-08-06T21:15:52.1960503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0128s] [ 87%] 2024-08-06T21:15:52.1961800Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.1753s] [ 87%] 2024-08-06T21:15:52.1963009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0161s] [ 87%] 2024-08-06T21:15:52.1964237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0163s] [ 88%] 2024-08-06T21:15:52.1965447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0113s] [ 88%] 2024-08-06T21:15:52.1966692Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0112s] [ 88%] 2024-08-06T21:15:52.1967928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0216s] [ 88%] 2024-08-06T21:15:52.1969143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0216s] [ 88%] 2024-08-06T21:15:52.1970417Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0130s] [ 88%] 2024-08-06T21:15:52.1971691Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0131s] [ 88%] 2024-08-06T21:15:52.1972919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0174s] [ 88%] 2024-08-06T21:15:52.1974145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0178s] [ 88%] 2024-08-06T21:15:52.1975384Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0125s] [ 88%] 2024-08-06T21:15:52.1976632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0125s] [ 88%] 2024-08-06T21:15:52.1977851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0230s] [ 88%] 2024-08-06T21:15:52.1979135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0230s] [ 88%] 2024-08-06T21:15:52.1980407Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0142s] [ 88%] 2024-08-06T21:15:52.1981650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0141s] [ 88%] 2024-08-06T21:15:52.1982854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0178s] [ 88%] 2024-08-06T21:15:52.1984093Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0182s] [ 88%] 2024-08-06T21:15:52.1985303Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0129s] [ 88%] 2024-08-06T21:15:52.1986560Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0129s] [ 88%] 2024-08-06T21:15:52.1987772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0234s] [ 88%] 2024-08-06T21:15:52.1989031Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0235s] [ 88%] 2024-08-06T21:15:52.1990313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0145s] [ 88%] 2024-08-06T21:15:52.1991529Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0144s] [ 88%] 2024-08-06T21:15:52.1992763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0209s] [ 88%] 2024-08-06T21:15:52.1993986Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0214s] [ 88%] 2024-08-06T21:15:52.1995219Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0151s] [ 88%] 2024-08-06T21:15:52.1996462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0151s] [ 88%] 2024-08-06T21:15:52.1997756Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.1974s] [ 88%] 2024-08-06T21:15:52.1999034Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0264s] [ 88%] 2024-08-06T21:15:52.2000249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0168s] [ 88%] 2024-08-06T21:15:52.2001490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0168s] [ 88%] 2024-08-06T21:15:52.2002702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0211s] [ 88%] 2024-08-06T21:15:52.2004301Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0217s] [ 88%] 2024-08-06T21:15:52.2005601Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0155s] [ 88%] 2024-08-06T21:15:52.2006859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0155s] [ 88%] 2024-08-06T21:15:52.2008167Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0267s] [ 88%] 2024-08-06T21:15:52.2009483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0267s] [ 88%] 2024-08-06T21:15:52.2010694Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0172s] [ 88%] 2024-08-06T21:15:52.2011911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0173s] [ 88%] 2024-08-06T21:15:52.2013150Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0148s] [ 88%] 2024-08-06T21:15:52.2014370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0149s] [ 88%] 2024-08-06T21:15:52.2015600Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0104s] [ 88%] 2024-08-06T21:15:52.2016903Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0103s] [ 88%] 2024-08-06T21:15:52.2018213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0204s] [ 88%] 2024-08-06T21:15:52.2019435Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0206s] [ 88%] 2024-08-06T21:15:52.2020667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0123s] [ 88%] 2024-08-06T21:15:52.2021897Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0122s] [ 88%] 2024-08-06T21:15:52.2023108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0151s] [ 88%] 2024-08-06T21:15:52.2024339Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0153s] [ 88%] 2024-08-06T21:15:52.2025543Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0108s] [ 88%] 2024-08-06T21:15:52.2026835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0107s] [ 88%] 2024-08-06T21:15:52.2028101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0207s] [ 88%] 2024-08-06T21:15:52.2029329Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0208s] [ 88%] 2024-08-06T21:15:52.2030540Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0126s] [ 88%] 2024-08-06T21:15:52.2031783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.1818s] [ 88%] 2024-08-06T21:15:52.2032995Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0128s] [ 88%] 2024-08-06T21:15:52.2034210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0115s] [ 88%] 2024-08-06T21:15:52.2035481Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0082s] [ 88%] 2024-08-06T21:15:52.2036770Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 88%] 2024-08-06T21:15:52.2038001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0157s] [ 88%] 2024-08-06T21:15:52.2039221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0157s] [ 88%] 2024-08-06T21:15:52.2040454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0090s] [ 88%] 2024-08-06T21:15:52.2041681Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0090s] [ 88%] 2024-08-06T21:15:52.2042900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0117s] [ 88%] 2024-08-06T21:15:52.2044109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0116s] [ 88%] 2024-08-06T21:15:52.2045358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0084s] [ 88%] 2024-08-06T21:15:52.2046657Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 88%] 2024-08-06T21:15:52.2047874Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0160s] [ 88%] 2024-08-06T21:15:52.2049111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0159s] [ 88%] 2024-08-06T21:15:52.2050323Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0093s] [ 88%] 2024-08-06T21:15:52.2051557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0093s] [ 88%] 2024-08-06T21:15:52.2052768Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0135s] [ 88%] 2024-08-06T21:15:52.2054051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0135s] [ 88%] 2024-08-06T21:15:52.2055313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0092s] [ 88%] 2024-08-06T21:15:52.2056557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0091s] [ 88%] 2024-08-06T21:15:52.2057788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0172s] [ 88%] 2024-08-06T21:15:52.2059012Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0175s] [ 88%] 2024-08-06T21:15:52.2060248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0106s] [ 89%] 2024-08-06T21:15:52.2061468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0098s] [ 89%] 2024-08-06T21:15:52.2062685Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0136s] [ 89%] 2024-08-06T21:15:52.2063939Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0136s] [ 89%] 2024-08-06T21:15:52.2065232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0092s] [ 89%] 2024-08-06T21:15:52.2066454Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0092s] [ 89%] 2024-08-06T21:15:52.2067670Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0174s] [ 89%] 2024-08-06T21:15:52.2068914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0174s] [ 89%] 2024-08-06T21:15:52.2070125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0102s] [ 89%] 2024-08-06T21:15:52.2071356Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0094s] [ 89%] 2024-08-06T21:15:52.2072628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0167s] [ 89%] 2024-08-06T21:15:52.2073929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0167s] [ 89%] 2024-08-06T21:15:52.2075145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0114s] [ 89%] 2024-08-06T21:15:52.2076386Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0114s] [ 89%] 2024-08-06T21:15:52.2077608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0205s] [ 89%] 2024-08-06T21:15:52.2078836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0205s] [ 89%] 2024-08-06T21:15:52.2080069Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0120s] [ 89%] 2024-08-06T21:15:52.2081290Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0120s] [ 89%] 2024-08-06T21:15:52.2082555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0168s] [ 89%] 2024-08-06T21:15:52.2083820Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0169s] [ 89%] 2024-08-06T21:15:52.2085043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0115s] [ 89%] 2024-08-06T21:15:52.2086262Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0115s] [ 89%] 2024-08-06T21:15:52.2087504Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0206s] [ 89%] 2024-08-06T21:15:52.2088726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0205s] [ 89%] 2024-08-06T21:15:52.2089937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0122s] [ 89%] 2024-08-06T21:15:52.2091214Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0123s] [ 89%] 2024-08-06T21:15:52.2092482Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0103s] [ 89%] 2024-08-06T21:15:52.2093719Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 89%] 2024-08-06T21:15:52.2094925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0073s] [ 89%] 2024-08-06T21:15:52.2096184Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 89%] 2024-08-06T21:15:52.2097404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0141s] [ 89%] 2024-08-06T21:15:52.2098643Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0141s] [ 89%] 2024-08-06T21:15:52.2099852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0080s] [ 89%] 2024-08-06T21:15:52.2101115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0081s] [ 89%] 2024-08-06T21:15:52.2102394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0104s] [ 89%] 2024-08-06T21:15:52.2103608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0104s] [ 89%] 2024-08-06T21:15:52.2105279Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0075s] [ 89%] 2024-08-06T21:15:52.2106524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 89%] 2024-08-06T21:15:52.2107773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0144s] [ 89%] 2024-08-06T21:15:52.2108981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0143s] [ 89%] 2024-08-06T21:15:52.2110289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0083s] [ 89%] 2024-08-06T21:15:52.2111578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_512_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 89%] 2024-08-06T21:15:52.2112794Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 89%] 2024-08-06T21:15:52.2114029Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 89%] 2024-08-06T21:15:52.2115241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 89%] 2024-08-06T21:15:52.2116486Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 89%] 2024-08-06T21:15:52.2117713Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0084s] [ 89%] 2024-08-06T21:15:52.2118958Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0087s] [ 89%] 2024-08-06T21:15:52.2120231Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 89%] 2024-08-06T21:15:52.2121539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 89%] 2024-08-06T21:15:52.2122743Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 89%] 2024-08-06T21:15:52.2123972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 89%] 2024-08-06T21:15:52.2125182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 89%] 2024-08-06T21:15:52.2126409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 89%] 2024-08-06T21:15:52.2127650Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 89%] 2024-08-06T21:15:52.2128911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0078s] [ 89%] 2024-08-06T21:15:52.2130193Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 89%] 2024-08-06T21:15:52.2131416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 89%] 2024-08-06T21:15:52.2132647Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0072s] [ 89%] 2024-08-06T21:15:52.2133864Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 89%] 2024-08-06T21:15:52.2135101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0064s] [ 89%] 2024-08-06T21:15:52.2136321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 89%] 2024-08-06T21:15:52.2137547Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0086s] [ 89%] 2024-08-06T21:15:52.2138838Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0089s] [ 89%] 2024-08-06T21:15:52.2140114Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0070s] [ 89%] 2024-08-06T21:15:52.2141354Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0065s] [ 89%] 2024-08-06T21:15:52.2142558Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0062s] [ 89%] 2024-08-06T21:15:52.2143793Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 89%] 2024-08-06T21:15:52.2145000Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 89%] 2024-08-06T21:15:52.2146227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 89%] 2024-08-06T21:15:52.2147493Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0089s] [ 89%] 2024-08-06T21:15:52.2148763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0081s] [ 89%] 2024-08-06T21:15:52.2149993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0078s] [ 89%] 2024-08-06T21:15:52.2151205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 89%] 2024-08-06T21:15:52.2152437Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0087s] [ 89%] 2024-08-06T21:15:52.2153663Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0094s] [ 89%] 2024-08-06T21:15:52.2154893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0074s] [ 89%] 2024-08-06T21:15:52.2156111Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 90%] 2024-08-06T21:15:52.2157404Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0106s] [ 90%] 2024-08-06T21:15:52.2158679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0107s] [ 90%] 2024-08-06T21:15:52.2159896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0087s] [ 90%] 2024-08-06T21:15:52.2161136Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0086s] [ 90%] 2024-08-06T21:15:52.2162345Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0080s] [ 90%] 2024-08-06T21:15:52.2163584Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0078s] [ 90%] 2024-08-06T21:15:52.2164786Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0074s] [ 90%] 2024-08-06T21:15:52.2166059Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 90%] 2024-08-06T21:15:52.2167331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0085s] [ 90%] 2024-08-06T21:15:52.2168569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0091s] [ 90%] 2024-08-06T21:15:52.2169780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0076s] [ 90%] 2024-08-06T21:15:52.2170994Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 90%] 2024-08-06T21:15:52.2172224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 90%] 2024-08-06T21:15:52.2173443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 90%] 2024-08-06T21:15:52.2174667Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 90%] 2024-08-06T21:15:52.2175927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 90%] 2024-08-06T21:15:52.2177228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 90%] 2024-08-06T21:15:52.2178448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 90%] 2024-08-06T21:15:52.2179674Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 90%] 2024-08-06T21:15:52.2180901Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 90%] 2024-08-06T21:15:52.2182113Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0060s] [ 90%] 2024-08-06T21:15:52.2183342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 90%] 2024-08-06T21:15:52.2184595Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 90%] 2024-08-06T21:15:52.2185879Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 90%] 2024-08-06T21:15:52.2187107Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0072s] [ 90%] 2024-08-06T21:15:52.2188335Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 90%] 2024-08-06T21:15:52.2189555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0064s] [ 90%] 2024-08-06T21:15:52.2190839Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 90%] 2024-08-06T21:15:52.2192060Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0110s] [ 90%] 2024-08-06T21:15:52.2193275Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0115s] [ 90%] 2024-08-06T21:15:52.2194550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0084s] [ 90%] 2024-08-06T21:15:52.2195840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 90%] 2024-08-06T21:15:52.2197092Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0138s] [ 90%] 2024-08-06T21:15:52.2198316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0139s] [ 90%] 2024-08-06T21:15:52.2199554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0093s] [ 90%] 2024-08-06T21:15:52.2200788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0091s] [ 90%] 2024-08-06T21:15:52.2202015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0095s] [ 90%] 2024-08-06T21:15:52.2203270Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0093s] [ 90%] 2024-08-06T21:15:52.2204980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0080s] [ 90%] 2024-08-06T21:15:52.2206247Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 90%] 2024-08-06T21:15:52.2207480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0107s] [ 90%] 2024-08-06T21:15:52.2208720Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0097s] [ 90%] 2024-08-06T21:15:52.2209936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0083s] [ 90%] 2024-08-06T21:15:52.2211174Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0084s] [ 90%] 2024-08-06T21:15:52.2212385Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0115s] [ 90%] 2024-08-06T21:15:52.2213704Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0125s] [ 90%] 2024-08-06T21:15:52.2214991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0089s] [ 90%] 2024-08-06T21:15:52.2216212Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0088s] [ 90%] 2024-08-06T21:15:52.2217466Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0147s] [ 90%] 2024-08-06T21:15:52.2218697Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0149s] [ 90%] 2024-08-06T21:15:52.2219941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0101s] [ 90%] 2024-08-06T21:15:52.2221164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0104s] [ 90%] 2024-08-06T21:15:52.2222446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0101s] [ 90%] 2024-08-06T21:15:52.2223727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 90%] 2024-08-06T21:15:52.2224963Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0091s] [ 90%] 2024-08-06T21:15:52.2226179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0092s] [ 90%] 2024-08-06T21:15:52.2227411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0108s] [ 90%] 2024-08-06T21:15:52.2228654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0104s] [ 90%] 2024-08-06T21:15:52.2229865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0089s] [ 90%] 2024-08-06T21:15:52.2231098Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0092s] [ 90%] 2024-08-06T21:15:52.2232400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0128s] [ 90%] 2024-08-06T21:15:52.2233703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0131s] [ 90%] 2024-08-06T21:15:52.2234914Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0093s] [ 90%] 2024-08-06T21:15:52.2236147Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0094s] [ 90%] 2024-08-06T21:15:52.2237398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0156s] [ 90%] 2024-08-06T21:15:52.2238631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0157s] [ 90%] 2024-08-06T21:15:52.2239870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0106s] [ 90%] 2024-08-06T21:15:52.2241141Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0106s] [ 90%] 2024-08-06T21:15:52.2242420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0104s] [ 90%] 2024-08-06T21:15:52.2243634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0105s] [ 90%] 2024-08-06T21:15:52.2244865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0093s] [ 90%] 2024-08-06T21:15:52.2246078Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0094s] [ 90%] 2024-08-06T21:15:52.2247338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0120s] [ 90%] 2024-08-06T21:15:52.2248555Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0120s] [ 90%] 2024-08-06T21:15:52.2249766Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0107s] [ 90%] 2024-08-06T21:15:52.2251049Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0105s] [ 90%] 2024-08-06T21:15:52.2252313Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0096s] [ 90%] 2024-08-06T21:15:52.2253551Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0099s] [ 91%] 2024-08-06T21:15:52.2254763Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0069s] [ 91%] 2024-08-06T21:15:52.2256005Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 91%] 2024-08-06T21:15:52.2257244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0124s] [ 91%] 2024-08-06T21:15:52.2258490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0125s] [ 91%] 2024-08-06T21:15:52.2259701Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0080s] [ 91%] 2024-08-06T21:15:52.2260968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 91%] 2024-08-06T21:15:52.2262251Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0079s] [ 91%] 2024-08-06T21:15:52.2263460Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0079s] [ 91%] 2024-08-06T21:15:52.2264679Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0072s] [ 91%] 2024-08-06T21:15:52.2265900Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0071s] [ 91%] 2024-08-06T21:15:52.2267159Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0093s] [ 91%] 2024-08-06T21:15:52.2268374Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0093s] [ 91%] 2024-08-06T21:15:52.2269646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0081s] [ 91%] 2024-08-06T21:15:52.2270916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0081s] [ 91%] 2024-08-06T21:15:52.2272133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 91%] 2024-08-06T21:15:52.2273364Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 91%] 2024-08-06T21:15:52.2274573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 91%] 2024-08-06T21:15:52.2275815Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0049s] [ 91%] 2024-08-06T21:15:52.2277048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 91%] 2024-08-06T21:15:52.2278281Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 91%] 2024-08-06T21:15:52.2279546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 91%] 2024-08-06T21:15:52.2280836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 91%] 2024-08-06T21:15:52.2282038Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 91%] 2024-08-06T21:15:52.2283244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 91%] 2024-08-06T21:15:52.2284471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 91%] 2024-08-06T21:15:52.2285678Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 91%] 2024-08-06T21:15:52.2286920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 91%] 2024-08-06T21:15:52.2288134Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 91%] 2024-08-06T21:15:52.2289395Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 91%] 2024-08-06T21:15:52.2290653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 91%] 2024-08-06T21:15:52.2291880Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 91%] 2024-08-06T21:15:52.2293099Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 91%] 2024-08-06T21:15:52.2294311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 91%] 2024-08-06T21:15:52.2295539Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 91%] 2024-08-06T21:15:52.2296771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 91%] 2024-08-06T21:15:52.2298050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 91%] 2024-08-06T21:15:52.2299316Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 91%] 2024-08-06T21:15:52.2300549Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 91%] 2024-08-06T21:15:52.2301750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 91%] 2024-08-06T21:15:52.2302983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 91%] 2024-08-06T21:15:52.2304590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 91%] 2024-08-06T21:15:52.2305858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 91%] 2024-08-06T21:15:52.2307102Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 91%] 2024-08-06T21:15:52.2308394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 91%] 2024-08-06T21:15:52.2309682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 91%] 2024-08-06T21:15:52.2310889Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 91%] 2024-08-06T21:15:52.2312115Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 91%] 2024-08-06T21:15:52.2313333Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 91%] 2024-08-06T21:15:52.2314557Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 91%] 2024-08-06T21:15:52.2315767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 91%] 2024-08-06T21:15:52.2317042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 91%] 2024-08-06T21:15:52.2318349Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 91%] 2024-08-06T21:15:52.2319563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0069s] [ 91%] 2024-08-06T21:15:52.2320791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 91%] 2024-08-06T21:15:52.2321989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0065s] [ 91%] 2024-08-06T21:15:52.2323215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 91%] 2024-08-06T21:15:52.2324411Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0062s] [ 91%] 2024-08-06T21:15:52.2325633Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 91%] 2024-08-06T21:15:52.2326896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0075s] [ 91%] 2024-08-06T21:15:52.2328158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 91%] 2024-08-06T21:15:52.2329377Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 91%] 2024-08-06T21:15:52.2330587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0067s] [ 91%] 2024-08-06T21:15:52.2331819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 91%] 2024-08-06T21:15:52.2333028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 91%] 2024-08-06T21:15:52.2334244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 91%] 2024-08-06T21:15:52.2335494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 91%] 2024-08-06T21:15:52.2336767Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0057s] [ 91%] 2024-08-06T21:15:52.2337976Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 91%] 2024-08-06T21:15:52.2339179Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0055s] [ 91%] 2024-08-06T21:15:52.2340414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 91%] 2024-08-06T21:15:52.2341615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 91%] 2024-08-06T21:15:52.2342832Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 91%] 2024-08-06T21:15:52.2344026Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0059s] [ 91%] 2024-08-06T21:15:52.2345289Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 91%] 2024-08-06T21:15:52.2346553Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 91%] 2024-08-06T21:15:52.2347790Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 91%] 2024-08-06T21:15:52.2348991Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 92%] 2024-08-06T21:15:52.2350197Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 92%] 2024-08-06T21:15:52.2351433Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0178s] [ 92%] 2024-08-06T21:15:52.2352649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0184s] [ 92%] 2024-08-06T21:15:52.2353929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0113s] [ 92%] 2024-08-06T21:15:52.2355205Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0111s] [ 92%] 2024-08-06T21:15:52.2356449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0225s] [ 92%] 2024-08-06T21:15:52.2357686Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0240s] [ 92%] 2024-08-06T21:15:52.2358925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0142s] [ 92%] 2024-08-06T21:15:52.2360153Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0139s] [ 92%] 2024-08-06T21:15:52.2361361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0121s] [ 92%] 2024-08-06T21:15:52.2362589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0129s] [ 92%] 2024-08-06T21:15:52.2363841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0119s] [ 92%] 2024-08-06T21:15:52.2365128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0121s] [ 92%] 2024-08-06T21:15:52.2366341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0133s] [ 92%] 2024-08-06T21:15:52.2367590Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0144s] [ 92%] 2024-08-06T21:15:52.2368814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0130s] [ 92%] 2024-08-06T21:15:52.2370051Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0134s] [ 92%] 2024-08-06T21:15:52.2371265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0192s] [ 92%] 2024-08-06T21:15:52.2372525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.1952s] [ 92%] 2024-08-06T21:15:52.2373811Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0135s] [ 92%] 2024-08-06T21:15:52.2375028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0136s] [ 92%] 2024-08-06T21:15:52.2376264Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0255s] [ 92%] 2024-08-06T21:15:52.2377503Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0248s] [ 92%] 2024-08-06T21:15:52.2378750Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0156s] [ 92%] 2024-08-06T21:15:52.2379971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0156s] [ 92%] 2024-08-06T21:15:52.2381191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0144s] [ 92%] 2024-08-06T21:15:52.2382450Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0146s] [ 92%] 2024-08-06T21:15:52.2383711Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0122s] [ 92%] 2024-08-06T21:15:52.2384945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0134s] [ 92%] 2024-08-06T21:15:52.2386160Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0162s] [ 92%] 2024-08-06T21:15:52.2387419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0170s] [ 92%] 2024-08-06T21:15:52.2388632Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0157s] [ 92%] 2024-08-06T21:15:52.2389865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0156s] [ 92%] 2024-08-06T21:15:52.2391119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0233s] [ 92%] 2024-08-06T21:15:52.2392416Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0232s] [ 92%] 2024-08-06T21:15:52.2393625Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0164s] [ 92%] 2024-08-06T21:15:52.2394844Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0161s] [ 92%] 2024-08-06T21:15:52.2396086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0285s] [ 92%] 2024-08-06T21:15:52.2397331Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0293s] [ 92%] 2024-08-06T21:15:52.2398564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0176s] [ 92%] 2024-08-06T21:15:52.2399788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0204s] [ 92%] 2024-08-06T21:15:52.2401055Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0191s] [ 92%] 2024-08-06T21:15:52.2402322Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0191s] [ 92%] 2024-08-06T21:15:52.2403546Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.1841s] [ 92%] 2024-08-06T21:15:52.2405192Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0171s] [ 92%] 2024-08-06T21:15:52.2406442Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0203s] [ 92%] 2024-08-06T21:15:52.2407698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0203s] [ 92%] 2024-08-06T21:15:52.2408907Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0190s] [ 92%] 2024-08-06T21:15:52.2410216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0192s] [ 92%] 2024-08-06T21:15:52.2411507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0174s] [ 92%] 2024-08-06T21:15:52.2412746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0179s] [ 92%] 2024-08-06T21:15:52.2413954Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0120s] [ 92%] 2024-08-06T21:15:52.2415198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0120s] [ 92%] 2024-08-06T21:15:52.2416424Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0224s] [ 92%] 2024-08-06T21:15:52.2417664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0230s] [ 92%] 2024-08-06T21:15:52.2418899Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0137s] [ 92%] 2024-08-06T21:15:52.2420198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0131s] [ 92%] 2024-08-06T21:15:52.2421494Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0125s] [ 92%] 2024-08-06T21:15:52.2422705Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0125s] [ 92%] 2024-08-06T21:15:52.2423935Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0113s] [ 92%] 2024-08-06T21:15:52.2425157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0113s] [ 92%] 2024-08-06T21:15:52.2426391Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0140s] [ 92%] 2024-08-06T21:15:52.2427623Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0139s] [ 92%] 2024-08-06T21:15:52.2428873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0129s] [ 92%] 2024-08-06T21:15:52.2430851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0126s] [ 92%] 2024-08-06T21:15:52.2432071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 92%] 2024-08-06T21:15:52.2433306Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 92%] 2024-08-06T21:15:52.2434514Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 92%] 2024-08-06T21:15:52.2435757Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 92%] 2024-08-06T21:15:52.2436968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0069s] [ 92%] 2024-08-06T21:15:52.2438227Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 92%] 2024-08-06T21:15:52.2439487Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0066s] [ 92%] 2024-08-06T21:15:52.2440761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 92%] 2024-08-06T21:15:52.2441979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0061s] [ 92%] 2024-08-06T21:15:52.2443191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 92%] 2024-08-06T21:15:52.2444420Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 92%] 2024-08-06T21:15:52.2445628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 92%] 2024-08-06T21:15:52.2446854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0065s] [ 93%] 2024-08-06T21:15:52.2448135Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 93%] 2024-08-06T21:15:52.2449419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 93%] 2024-08-06T21:15:52.2450638Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 93%] 2024-08-06T21:15:52.2451849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 93%] 2024-08-06T21:15:52.2453086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 93%] 2024-08-06T21:15:52.2454300Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 93%] 2024-08-06T21:15:52.2455538Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 93%] 2024-08-06T21:15:52.2456773Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 93%] 2024-08-06T21:15:52.2458086Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 93%] 2024-08-06T21:15:52.2459351Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 93%] 2024-08-06T21:15:52.2460589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 93%] 2024-08-06T21:15:52.2461789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 93%] 2024-08-06T21:15:52.2463006Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 93%] 2024-08-06T21:15:52.2464232Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 93%] 2024-08-06T21:15:52.2465499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 93%] 2024-08-06T21:15:52.2466836Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0069s] [ 93%] 2024-08-06T21:15:52.2476163Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0068s] [ 93%] 2024-08-06T21:15:52.2477509Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 93%] 2024-08-06T21:15:52.2478728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 93%] 2024-08-06T21:15:52.2479944Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0063s] [ 93%] 2024-08-06T21:15:52.2481158Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 93%] 2024-08-06T21:15:52.2482357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0067s] [ 93%] 2024-08-06T21:15:52.2483559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 93%] 2024-08-06T21:15:52.2484865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0084s] [ 93%] 2024-08-06T21:15:52.2486166Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 93%] 2024-08-06T21:15:52.2487378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0079s] [ 93%] 2024-08-06T21:15:52.2488586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 93%] 2024-08-06T21:15:52.2489842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0074s] [ 93%] 2024-08-06T21:15:52.2491070Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 93%] 2024-08-06T21:15:52.2492259Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0068s] [ 93%] 2024-08-06T21:15:52.2493510Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 93%] 2024-08-06T21:15:52.2494758Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0087s] [ 93%] 2024-08-06T21:15:52.2495987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0086s] [ 93%] 2024-08-06T21:15:52.2497181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0077s] [ 93%] 2024-08-06T21:15:52.2498394Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 93%] 2024-08-06T21:15:52.2499589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 93%] 2024-08-06T21:15:52.2500785Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 93%] 2024-08-06T21:15:52.2501981Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 93%] 2024-08-06T21:15:52.2503221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 93%] 2024-08-06T21:15:52.2504987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 93%] 2024-08-06T21:15:52.2506246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 93%] 2024-08-06T21:15:52.2507444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 93%] 2024-08-06T21:15:52.2508654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 93%] 2024-08-06T21:15:52.2509855Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 93%] 2024-08-06T21:15:52.2511042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 93%] 2024-08-06T21:15:52.2512324Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0057s] [ 93%] 2024-08-06T21:15:52.2513649Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 93%] 2024-08-06T21:15:52.2514841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0071s] [ 93%] 2024-08-06T21:15:52.2516053Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0069s] [ 93%] 2024-08-06T21:15:52.2517244Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 93%] 2024-08-06T21:15:52.2518445Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 93%] 2024-08-06T21:15:52.2519640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 93%] 2024-08-06T21:15:52.2520841Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 93%] 2024-08-06T21:15:52.2522090Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 93%] 2024-08-06T21:15:52.2523355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 93%] 2024-08-06T21:15:52.2524554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 93%] 2024-08-06T21:15:52.2525771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 93%] 2024-08-06T21:15:52.2526978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 93%] 2024-08-06T21:15:52.2528189Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 93%] 2024-08-06T21:15:52.2529375Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 93%] 2024-08-06T21:15:52.2530565Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 93%] 2024-08-06T21:15:52.2531801Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 93%] 2024-08-06T21:15:52.2533048Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 93%] 2024-08-06T21:15:52.2534238Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 93%] 2024-08-06T21:15:52.2535446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 93%] 2024-08-06T21:15:52.2536644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 93%] 2024-08-06T21:15:52.2537851Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 93%] 2024-08-06T21:15:52.2539041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 93%] 2024-08-06T21:15:52.2540288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 93%] 2024-08-06T21:15:52.2541532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 93%] 2024-08-06T21:15:52.2542736Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0049s] [ 93%] 2024-08-06T21:15:52.2543933Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 93%] 2024-08-06T21:15:52.2545149Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 93%] 2024-08-06T21:15:52.2546358Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 93%] 2024-08-06T21:15:52.2547559Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 93%] 2024-08-06T21:15:52.2548749Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 93%] 2024-08-06T21:15:52.2549999Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 94%] 2024-08-06T21:15:52.2551245Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 94%] 2024-08-06T21:15:52.2552436Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 94%] 2024-08-06T21:15:52.2553636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 94%] 2024-08-06T21:15:52.2554835Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 94%] 2024-08-06T21:15:52.2556037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 94%] 2024-08-06T21:15:52.2557265Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 94%] 2024-08-06T21:15:52.2558468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 94%] 2024-08-06T21:15:52.2559723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 94%] 2024-08-06T21:15:52.2560967Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 94%] 2024-08-06T21:15:52.2562164Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 94%] 2024-08-06T21:15:52.2563361Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 94%] 2024-08-06T21:15:52.2564583Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 94%] 2024-08-06T21:15:52.2565789Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 94%] 2024-08-06T21:15:52.2566992Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 94%] 2024-08-06T21:15:52.2568221Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 94%] 2024-08-06T21:15:52.2569463Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 94%] 2024-08-06T21:15:52.2570653Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 94%] 2024-08-06T21:15:52.2571840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 94%] 2024-08-06T21:15:52.2573043Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 94%] 2024-08-06T21:15:52.2574240Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 94%] 2024-08-06T21:15:52.2575447Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 94%] 2024-08-06T21:15:52.2576645Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 94%] 2024-08-06T21:15:52.2577887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0049s] [ 94%] 2024-08-06T21:15:52.2579132Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0049s] [ 94%] 2024-08-06T21:15:52.2580321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 94%] 2024-08-06T21:15:52.2581522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0049s] [ 94%] 2024-08-06T21:15:52.2582723Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 94%] 2024-08-06T21:15:52.2583927Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 94%] 2024-08-06T21:15:52.2585137Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 94%] 2024-08-06T21:15:52.2586393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 94%] 2024-08-06T21:15:52.2587628Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 94%] 2024-08-06T21:15:52.2588817Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 94%] 2024-08-06T21:15:52.2590001Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 94%] 2024-08-06T21:15:52.2591199Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 94%] 2024-08-06T21:15:52.2592406Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 94%] 2024-08-06T21:15:52.2593592Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 94%] 2024-08-06T21:15:52.2594804Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 94%] 2024-08-06T21:15:52.2596036Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_64_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 94%] 2024-08-06T21:15:52.2597294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 94%] 2024-08-06T21:15:52.2598501Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 94%] 2024-08-06T21:15:52.2599698Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0048s] [ 94%] 2024-08-06T21:15:52.2600917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 94%] 2024-08-06T21:15:52.2602125Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 94%] 2024-08-06T21:15:52.2603338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 94%] 2024-08-06T21:15:52.2605095Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 94%] 2024-08-06T21:15:52.2606409Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 94%] 2024-08-06T21:15:52.2607603Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 94%] 2024-08-06T21:15:52.2608808Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 94%] 2024-08-06T21:15:52.2610007Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 94%] 2024-08-06T21:15:52.2611213Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 94%] 2024-08-06T21:15:52.2612415Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 94%] 2024-08-06T21:15:52.2613617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 94%] 2024-08-06T21:15:52.2614905Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 94%] 2024-08-06T21:15:52.2616198Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 94%] 2024-08-06T21:15:52.2617405Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 94%] 2024-08-06T21:15:52.2618612Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 94%] 2024-08-06T21:15:52.2619826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 94%] 2024-08-06T21:15:52.2621037Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0049s] [ 94%] 2024-08-06T21:15:52.2622241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 94%] 2024-08-06T21:15:52.2623502Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 94%] 2024-08-06T21:15:52.2624781Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 94%] 2024-08-06T21:15:52.2625997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 94%] 2024-08-06T21:15:52.2627191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 94%] 2024-08-06T21:15:52.2628402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 94%] 2024-08-06T21:15:52.2629597Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 94%] 2024-08-06T21:15:52.2630799Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 94%] 2024-08-06T21:15:52.2631997Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 94%] 2024-08-06T21:15:52.2633246Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 94%] 2024-08-06T21:15:52.2634520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0065s] [ 94%] 2024-08-06T21:15:52.2635722Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 94%] 2024-08-06T21:15:52.2636928Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0066s] [ 94%] 2024-08-06T21:15:52.2638143Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0066s] [ 94%] 2024-08-06T21:15:52.2639348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 94%] 2024-08-06T21:15:52.2640554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 94%] 2024-08-06T21:15:52.2641803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0070s] [ 94%] 2024-08-06T21:15:52.2643065Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 94%] 2024-08-06T21:15:52.2644294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 95%] 2024-08-06T21:15:52.2645507Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 95%] 2024-08-06T21:15:52.2646706Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 95%] 2024-08-06T21:15:52.2647916Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 95%] 2024-08-06T21:15:52.2649109Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 95%] 2024-08-06T21:15:52.2650311Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 95%] 2024-08-06T21:15:52.2651550Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 95%] 2024-08-06T21:15:52.2652816Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 95%] 2024-08-06T21:15:52.2654027Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 95%] 2024-08-06T21:15:52.2655237Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 95%] 2024-08-06T21:15:52.2656446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0050s] [ 95%] 2024-08-06T21:15:52.2657652Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 95%] 2024-08-06T21:15:52.2658854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0049s] [ 95%] 2024-08-06T21:15:52.2660100Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 95%] 2024-08-06T21:15:52.2661357Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0057s] [ 95%] 2024-08-06T21:15:52.2662566Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0058s] [ 95%] 2024-08-06T21:15:52.2663779Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0055s] [ 95%] 2024-08-06T21:15:52.2664993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 95%] 2024-08-06T21:15:52.2666191Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 95%] 2024-08-06T21:15:52.2667393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 95%] 2024-08-06T21:15:52.2668578Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 95%] 2024-08-06T21:15:52.2669823Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0055s] [ 95%] 2024-08-06T21:15:52.2671067Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 95%] 2024-08-06T21:15:52.2672272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0064s] [ 95%] 2024-08-06T21:15:52.2673472Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 95%] 2024-08-06T21:15:52.2674683Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_128_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 95%] 2024-08-06T21:15:52.2675888Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0068s] [ 95%] 2024-08-06T21:15:52.2677097Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0075s] [ 95%] 2024-08-06T21:15:52.2678342Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 95%] 2024-08-06T21:15:52.2679602Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0057s] [ 95%] 2024-08-06T21:15:52.2680814Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0085s] [ 95%] 2024-08-06T21:15:52.2682021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0084s] [ 95%] 2024-08-06T21:15:52.2683252Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 95%] 2024-08-06T21:15:52.2684468Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 95%] 2024-08-06T21:15:52.2685668Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0062s] [ 95%] 2024-08-06T21:15:52.2686863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 95%] 2024-08-06T21:15:52.2688110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 95%] 2024-08-06T21:15:52.2689370Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 95%] 2024-08-06T21:15:52.2690570Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0070s] [ 95%] 2024-08-06T21:15:52.2691780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 95%] 2024-08-06T21:15:52.2692983Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 95%] 2024-08-06T21:15:52.2694200Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 95%] 2024-08-06T21:15:52.2695398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0078s] [ 95%] 2024-08-06T21:15:52.2696646Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0083s] [ 95%] 2024-08-06T21:15:52.2697895Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 95%] 2024-08-06T21:15:52.2699104Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 95%] 2024-08-06T21:15:52.2700312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0089s] [ 95%] 2024-08-06T21:15:52.2701522Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0093s] [ 95%] 2024-08-06T21:15:52.2702739Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 95%] 2024-08-06T21:15:52.2704334Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0066s] [ 95%] 2024-08-06T21:15:52.2705614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0066s] [ 95%] 2024-08-06T21:15:52.2706908Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0067s] [ 95%] 2024-08-06T21:15:52.2708181Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 95%] 2024-08-06T21:15:52.2709378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 95%] 2024-08-06T21:15:52.2710575Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0071s] [ 95%] 2024-08-06T21:15:52.2711788Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 95%] 2024-08-06T21:15:52.2712989Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0071s] [ 95%] 2024-08-06T21:15:52.2714210Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0068s] [ 95%] 2024-08-06T21:15:52.2715462Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0104s] [ 95%] 2024-08-06T21:15:52.2716735Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0106s] [ 95%] 2024-08-06T21:15:52.2717936Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0070s] [ 95%] 2024-08-06T21:15:52.2719145Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0070s] [ 95%] 2024-08-06T21:15:52.2720348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0109s] [ 95%] 2024-08-06T21:15:52.2721563Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0110s] [ 95%] 2024-08-06T21:15:52.2722771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0080s] [ 95%] 2024-08-06T21:15:52.2723987Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0080s] [ 95%] 2024-08-06T21:15:52.2725234Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0079s] [ 95%] 2024-08-06T21:15:52.2726484Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0081s] [ 95%] 2024-08-06T21:15:52.2727682Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0071s] [ 95%] 2024-08-06T21:15:52.2728876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 95%] 2024-08-06T21:15:52.2730085Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0085s] [ 95%] 2024-08-06T21:15:52.2731294Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0086s] [ 95%] 2024-08-06T21:15:52.2732490Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0078s] [ 95%] 2024-08-06T21:15:52.2733710Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0077s] [ 95%] 2024-08-06T21:15:52.2734950Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0071s] [ 95%] 2024-08-06T21:15:52.2736208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0071s] [ 95%] 2024-08-06T21:15:52.2737400Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 95%] 2024-08-06T21:15:52.2738607Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 95%] 2024-08-06T21:15:52.2739810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0078s] [ 96%] 2024-08-06T21:15:52.2741021Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0077s] [ 96%] 2024-08-06T21:15:52.2742215Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 96%] 2024-08-06T21:15:52.2743480Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 96%] 2024-08-06T21:15:52.2744726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 96%] 2024-08-06T21:15:52.2745917Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0057s] [ 96%] 2024-08-06T21:15:52.2747108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 96%] 2024-08-06T21:15:52.2748304Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 96%] 2024-08-06T21:15:52.2749508Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 96%] 2024-08-06T21:15:52.2750703Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 96%] 2024-08-06T21:15:52.2751896Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 96%] 2024-08-06T21:15:52.2753157Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_256_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 96%] 2024-08-06T21:15:52.2754419Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 96%] 2024-08-06T21:15:52.2755617Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 96%] 2024-08-06T21:15:52.2756810Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 96%] 2024-08-06T21:15:52.2758014Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 96%] 2024-08-06T21:15:52.2759208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 96%] 2024-08-06T21:15:52.2760414Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 96%] 2024-08-06T21:15:52.2761606Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 96%] 2024-08-06T21:15:52.2762881Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 96%] 2024-08-06T21:15:52.2764119Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 96%] 2024-08-06T21:15:52.2765305Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 96%] 2024-08-06T21:15:52.2766495Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 96%] 2024-08-06T21:15:52.2767693Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 96%] 2024-08-06T21:15:52.2768886Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 96%] 2024-08-06T21:15:52.2770077Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 96%] 2024-08-06T21:15:52.2771315Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 96%] 2024-08-06T21:15:52.2772581Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 96%] 2024-08-06T21:15:52.2773778Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 96%] 2024-08-06T21:15:52.2774972Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 96%] 2024-08-06T21:15:52.2776169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 96%] 2024-08-06T21:15:52.2777373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 96%] 2024-08-06T21:15:52.2778564Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 96%] 2024-08-06T21:15:52.2779772Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 96%] 2024-08-06T21:15:52.2781011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 96%] 2024-08-06T21:15:52.2782288Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 96%] 2024-08-06T21:15:52.2783470Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 96%] 2024-08-06T21:15:52.2784665Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 96%] 2024-08-06T21:15:52.2785850Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0054s] [ 96%] 2024-08-06T21:15:52.2787042Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 96%] 2024-08-06T21:15:52.2788236Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 96%] 2024-08-06T21:15:52.2789471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 96%] 2024-08-06T21:15:52.2790761Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 96%] 2024-08-06T21:15:52.2791968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 96%] 2024-08-06T21:15:52.2793161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 96%] 2024-08-06T21:15:52.2794355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 96%] 2024-08-06T21:15:52.2795554Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 96%] 2024-08-06T21:15:52.2796744Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 96%] 2024-08-06T21:15:52.2797938Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 96%] 2024-08-06T21:15:52.2799185Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 96%] 2024-08-06T21:15:52.2800432Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 96%] 2024-08-06T21:15:52.2801644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 96%] 2024-08-06T21:15:52.2802837Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 96%] 2024-08-06T21:15:52.2804418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 96%] 2024-08-06T21:15:52.2805672Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 96%] 2024-08-06T21:15:52.2806868Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 96%] 2024-08-06T21:15:52.2808182Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0064s] [ 96%] 2024-08-06T21:15:52.2809452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 96%] 2024-08-06T21:15:52.2810654Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 96%] 2024-08-06T21:15:52.2811842Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 96%] 2024-08-06T21:15:52.2813044Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0051s] [ 96%] 2024-08-06T21:15:52.2814241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 96%] 2024-08-06T21:15:52.2815428Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 96%] 2024-08-06T21:15:52.2816615Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 96%] 2024-08-06T21:15:52.2817870Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 96%] 2024-08-06T21:15:52.2819133Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 96%] 2024-08-06T21:15:52.2820326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 96%] 2024-08-06T21:15:52.2821526Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 96%] 2024-08-06T21:15:52.2822727Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 96%] 2024-08-06T21:15:52.2823930Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 96%] 2024-08-06T21:15:52.2825110Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 96%] 2024-08-06T21:15:52.2826355Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 96%] 2024-08-06T21:15:52.2827599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 96%] 2024-08-06T21:15:52.2828796Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 96%] 2024-08-06T21:15:52.2829979Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 96%] 2024-08-06T21:15:52.2831169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_4_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 96%] 2024-08-06T21:15:52.2832398Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0116s] [ 96%] 2024-08-06T21:15:52.2833608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0118s] [ 97%] 2024-08-06T21:15:52.2834812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0073s] [ 97%] 2024-08-06T21:15:52.2836066Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0075s] [ 97%] 2024-08-06T21:15:52.2837330Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0129s] [ 97%] 2024-08-06T21:15:52.2838541Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0129s] [ 97%] 2024-08-06T21:15:52.2839741Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0079s] [ 97%] 2024-08-06T21:15:52.2840961Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 97%] 2024-08-06T21:15:52.2842173Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0080s] [ 97%] 2024-08-06T21:15:52.2843378Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0080s] [ 97%] 2024-08-06T21:15:52.2844614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0074s] [ 97%] 2024-08-06T21:15:52.2845873Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0072s] [ 97%] 2024-08-06T21:15:52.2847074Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0086s] [ 97%] 2024-08-06T21:15:52.2848282Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0083s] [ 97%] 2024-08-06T21:15:52.2849478Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0081s] [ 97%] 2024-08-06T21:15:52.2850695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0079s] [ 97%] 2024-08-06T21:15:52.2851919Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0134s] [ 97%] 2024-08-06T21:15:52.2853124Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0136s] [ 97%] 2024-08-06T21:15:52.2854371Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0084s] [ 97%] 2024-08-06T21:15:52.2855634Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0082s] [ 97%] 2024-08-06T21:15:52.2856845Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0146s] [ 97%] 2024-08-06T21:15:52.2858050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.1888s] [ 97%] 2024-08-06T21:15:52.2859266Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0095s] [ 97%] 2024-08-06T21:15:52.2860476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0097s] [ 97%] 2024-08-06T21:15:52.2861695Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0093s] [ 97%] 2024-08-06T21:15:52.2862893Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0096s] [ 97%] 2024-08-06T21:15:52.2864130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0086s] [ 97%] 2024-08-06T21:15:52.2865393Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0083s] [ 97%] 2024-08-06T21:15:52.2866593Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0103s] [ 97%] 2024-08-06T21:15:52.2867809Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0103s] [ 97%] 2024-08-06T21:15:52.2869011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0094s] [ 97%] 2024-08-06T21:15:52.2870224Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0094s] [ 97%] 2024-08-06T21:15:52.2871446Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0168s] [ 97%] 2024-08-06T21:15:52.2872702Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0172s] [ 97%] 2024-08-06T21:15:52.2873952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0110s] [ 97%] 2024-08-06T21:15:52.2875161Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0111s] [ 97%] 2024-08-06T21:15:52.2876368Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0184s] [ 97%] 2024-08-06T21:15:52.2877586Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0184s] [ 97%] 2024-08-06T21:15:52.2878803Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0119s] [ 97%] 2024-08-06T21:15:52.2880011Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0120s] [ 97%] 2024-08-06T21:15:52.2881229Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0110s] [ 97%] 2024-08-06T21:15:52.2882473Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0113s] [ 97%] 2024-08-06T21:15:52.2883726Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0103s] [ 97%] 2024-08-06T21:15:52.2884925Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0103s] [ 97%] 2024-08-06T21:15:52.2886118Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0123s] [ 97%] 2024-08-06T21:15:52.2887332Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0123s] [ 97%] 2024-08-06T21:15:52.2888532Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0116s] [ 97%] 2024-08-06T21:15:52.2889737Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0115s] [ 97%] 2024-08-06T21:15:52.2890998Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0107s] [ 97%] 2024-08-06T21:15:52.2892272Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0111s] [ 97%] 2024-08-06T21:15:52.2893469Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0073s] [ 97%] 2024-08-06T21:15:52.2894675Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0074s] [ 97%] 2024-08-06T21:15:52.2895876Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0122s] [ 97%] 2024-08-06T21:15:52.2897091Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0119s] [ 97%] 2024-08-06T21:15:52.2898295Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0077s] [ 97%] 2024-08-06T21:15:52.2899499Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0077s] [ 97%] 2024-08-06T21:15:52.2900759Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0074s] [ 97%] 2024-08-06T21:15:52.2902009Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0072s] [ 97%] 2024-08-06T21:15:52.2903203Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0067s] [ 97%] 2024-08-06T21:15:52.2904812Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0069s] [ 97%] 2024-08-06T21:15:52.2906073Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0079s] [ 97%] 2024-08-06T21:15:52.2907277Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0078s] [ 97%] 2024-08-06T21:15:52.2908476Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0079s] [ 97%] 2024-08-06T21:15:52.2909689Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_512_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0078s] [ 97%] 2024-08-06T21:15:52.2910993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 97%] 2024-08-06T21:15:52.2912274Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 97%] 2024-08-06T21:15:52.2913471Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0056s] [ 97%] 2024-08-06T21:15:52.2914684Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 97%] 2024-08-06T21:15:52.2915887Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 97%] 2024-08-06T21:15:52.2917101Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 97%] 2024-08-06T21:15:52.2918298Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 97%] 2024-08-06T21:15:52.2919562Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 97%] 2024-08-06T21:15:52.2920849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 97%] 2024-08-06T21:15:52.2922050Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 97%] 2024-08-06T21:15:52.2923241Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 97%] 2024-08-06T21:15:52.2924444Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 97%] 2024-08-06T21:15:52.2925660Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0063s] [ 97%] 2024-08-06T21:15:52.2926859Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 97%] 2024-08-06T21:15:52.2928071Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 97%] 2024-08-06T21:15:52.2929320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 98%] 2024-08-06T21:15:52.2930573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 98%] 2024-08-06T21:15:52.2931783Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 98%] 2024-08-06T21:15:52.2932982Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 98%] 2024-08-06T21:15:52.2934202Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 98%] 2024-08-06T21:15:52.2935401Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 98%] 2024-08-06T21:15:52.2936614Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 98%] 2024-08-06T21:15:52.2937852Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 98%] 2024-08-06T21:15:52.2939121Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 98%] 2024-08-06T21:15:52.2940326Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 98%] 2024-08-06T21:15:52.2941525Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 98%] 2024-08-06T21:15:52.2942729Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 98%] 2024-08-06T21:15:52.2943937Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 98%] 2024-08-06T21:15:52.2945139Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0059s] [ 98%] 2024-08-06T21:15:52.2946338Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 98%] 2024-08-06T21:15:52.2947582Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 98%] 2024-08-06T21:15:52.2948829Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 98%] 2024-08-06T21:15:52.2950032Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 98%] 2024-08-06T21:15:52.2951249Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 98%] 2024-08-06T21:15:52.2952449Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0060s] [ 98%] 2024-08-06T21:15:52.2953666Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 98%] 2024-08-06T21:15:52.2954865Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0067s] [ 98%] 2024-08-06T21:15:52.2956130Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0065s] [ 98%] 2024-08-06T21:15:52.2957381Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 98%] 2024-08-06T21:15:52.2958599Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 98%] 2024-08-06T21:15:52.2959791Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0058s] [ 98%] 2024-08-06T21:15:52.2961015Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 98%] 2024-08-06T21:15:52.2962207Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0061s] [ 98%] 2024-08-06T21:15:52.2963402Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 98%] 2024-08-06T21:15:52.2964608Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0071s] [ 98%] 2024-08-06T21:15:52.2965862Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0070s] [ 98%] 2024-08-06T21:15:52.2967123Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0068s] [ 98%] 2024-08-06T21:15:52.2968320Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 98%] 2024-08-06T21:15:52.2969524Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0049s] [ 98%] 2024-08-06T21:15:52.2970740Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 98%] 2024-08-06T21:15:52.2971945Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 98%] 2024-08-06T21:15:52.2973138Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 98%] 2024-08-06T21:15:52.2974373Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 98%] 2024-08-06T21:15:52.2975636Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 98%] 2024-08-06T21:15:52.2976828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0057s] [ 98%] 2024-08-06T21:15:52.2978041Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 98%] 2024-08-06T21:15:52.2979228Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 98%] 2024-08-06T21:15:52.2980451Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0051s] [ 98%] 2024-08-06T21:15:52.2981631Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 98%] 2024-08-06T21:15:52.2982821Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 98%] 2024-08-06T21:15:52.2984079Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 98%] 2024-08-06T21:15:52.2985328Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 98%] 2024-08-06T21:15:52.2986523Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0062s] [ 98%] 2024-08-06T21:15:52.2987717Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_64_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 98%] 2024-08-06T21:15:52.2988920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0049s] [ 98%] 2024-08-06T21:15:52.2990144Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0049s] [ 98%] 2024-08-06T21:15:52.2991341Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0053s] [ 98%] 2024-08-06T21:15:52.2992573Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 98%] 2024-08-06T21:15:52.2993819Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0058s] [ 98%] 2024-08-06T21:15:52.2995035Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 98%] 2024-08-06T21:15:52.2996226Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0059s] [ 98%] 2024-08-06T21:15:52.2997438Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 98%] 2024-08-06T21:15:52.2998635Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 98%] 2024-08-06T21:15:52.2999849Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0056s] [ 98%] 2024-08-06T21:15:52.3001033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0050s] [ 98%] 2024-08-06T21:15:52.3002344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0053s] [ 98%] 2024-08-06T21:15:52.3003589Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0066s] [ 98%] 2024-08-06T21:15:52.3005216Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 98%] 2024-08-06T21:15:52.3006443Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0063s] [ 98%] 2024-08-06T21:15:52.3007644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_16_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0063s] [ 98%] 2024-08-06T21:15:52.3008854Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0055s] [ 98%] 2024-08-06T21:15:52.3010072Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 98%] 2024-08-06T21:15:52.3011348Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 98%] 2024-08-06T21:15:52.3012613Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0048s] [ 98%] 2024-08-06T21:15:52.3013826Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0057s] [ 98%] 2024-08-06T21:15:52.3015028Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0061s] [ 98%] 2024-08-06T21:15:52.3016230Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0056s] [ 98%] 2024-08-06T21:15:52.3017452Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0059s] [ 98%] 2024-08-06T21:15:52.3018644Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0054s] [ 98%] 2024-08-06T21:15:52.3019858Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 98%] 2024-08-06T21:15:52.3021108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 98%] 2024-08-06T21:15:52.3022376Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 98%] 2024-08-06T21:15:52.3023569Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 99%] 2024-08-06T21:15:52.3024771Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0059s] [ 99%] 2024-08-06T21:15:52.3025968Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 99%] 2024-08-06T21:15:52.3027172Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_32_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0062s] [ 99%] 2024-08-06T21:15:52.3028379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0056s] [ 99%] 2024-08-06T21:15:52.3029587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0054s] [ 99%] 2024-08-06T21:15:52.3030834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 99%] 2024-08-06T21:15:52.3032106Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0051s] [ 99%] 2024-08-06T21:15:52.3033312Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 99%] 2024-08-06T21:15:52.3034520Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 99%] 2024-08-06T21:15:52.3035734Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0060s] [ 99%] 2024-08-06T21:15:52.3036941Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0056s] [ 99%] 2024-08-06T21:15:52.3038128Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0057s] [ 99%] 2024-08-06T21:15:52.3039344Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0053s] [ 99%] 2024-08-06T21:15:52.3040580Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0055s] [ 99%] 2024-08-06T21:15:52.3041834Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0050s] [ 99%] 2024-08-06T21:15:52.3043033Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0060s] [ 99%] 2024-08-06T21:15:52.3044242Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0063s] [ 99%] 2024-08-06T21:15:52.3045441Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 99%] 2024-08-06T21:15:52.3046640Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_64_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0061s] [ 99%] 2024-08-06T21:15:52.3047828Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0052s] [ 99%] 2024-08-06T21:15:52.3049062Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0052s] [ 99%] 2024-08-06T21:15:52.3050321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0051s] [ 99%] 2024-08-06T21:15:52.3051517Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0052s] [ 99%] 2024-08-06T21:15:52.3052718Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0061s] [ 99%] 2024-08-06T21:15:52.3053920Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0060s] [ 99%] 2024-08-06T21:15:52.3055126Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0058s] [ 99%] 2024-08-06T21:15:52.3056321Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_False_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0058s] [ 99%] 2024-08-06T21:15:52.3057513Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale0_cuda_float16 PASSED [0.0053s] [ 99%] 2024-08-06T21:15:52.3058746Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float16_scale_l1_cuda_float16 PASSED [0.0055s] [ 99%] 2024-08-06T21:15:52.3059996Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale0_cuda_float32 PASSED [0.0052s] [ 99%] 2024-08-06T21:15:52.3061190Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_0_float32_scale_l1_cuda_float32 PASSED [0.0054s] [ 99%] 2024-08-06T21:15:52.3062379Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale0_cuda_float16 PASSED [0.0062s] [ 99%] 2024-08-06T21:15:52.3063587Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float16_scale_l1_cuda_float16 PASSED [0.0062s] [ 99%] 2024-08-06T21:15:52.3064776Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale0_cuda_float32 PASSED [0.0061s] [ 99%] 2024-08-06T21:15:52.3065975Z test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_vs_math_ref_grads_batch_size_8_seq_len_q_8_seq_len_k_8_head_dim_8_is_causal_True_dropout_p_0_22_float32_scale_l1_cuda_float32 PASSED [0.0060s] [ 99%] 2024-08-06T21:15:52.3066971Z test_transformers.py::TestSDPACudaOnlyCUDA::test_scaled_dot_product_attention_fused_kernels_packed_accuracy_type_dense_fused_kernel0_cuda PASSED [0.0271s] [ 99%] 2024-08-06T21:15:52.3067980Z test_transformers.py::TestSDPACudaOnlyCUDA::test_scaled_dot_product_attention_fused_kernels_packed_accuracy_type_nested_fused_kernel0_cuda PASSED [0.0469s] [ 99%] 2024-08-06T21:15:52.3068911Z test_transformers.py::TestSDPACudaOnlyCUDA::test_scaled_dot_product_attention_fused_kernels_packed_type_dense_is_contiguous_False_cuda PASSED [0.0083s] [ 99%] 2024-08-06T21:15:52.3069840Z test_transformers.py::TestSDPACudaOnlyCUDA::test_scaled_dot_product_attention_fused_kernels_packed_type_dense_is_contiguous_True_cuda PASSED [0.0081s] [ 99%] 2024-08-06T21:15:52.3070780Z test_transformers.py::TestSDPACudaOnlyCUDA::test_scaled_dot_product_attention_fused_kernels_packed_type_nested_is_contiguous_False_cuda PASSED [0.0324s] [ 99%] 2024-08-06T21:15:52.3071707Z test_transformers.py::TestSDPACudaOnlyCUDA::test_scaled_dot_product_attention_fused_kernels_packed_type_nested_is_contiguous_True_cuda PASSED [0.0297s] [ 99%] 2024-08-06T21:15:52.3072448Z test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_choice_with_determinism_warn_only_False_cuda PASSED [0.0014s] [ 99%] 2024-08-06T21:15:52.3073169Z test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_choice_with_determinism_warn_only_True_cuda PASSED [0.0013s] [ 99%] 2024-08-06T21:15:52.3074418Z test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_False_is_causal_False_bfloat16_cuda_bfloat16 SKIPPED [0.0003s] (Flash Attention was not built for this system) [ 99%] 2024-08-06T21:15:52.3075664Z test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_False_is_causal_False_float16_cuda_float16 SKIPPED [0.0002s] (Flash Attention was not built for this system) [ 99%] 2024-08-06T21:15:52.3076952Z test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_False_is_causal_True_bfloat16_cuda_bfloat16 SKIPPED [0.0003s] (Flash Attention was not built for this system) [ 99%] 2024-08-06T21:15:52.3078248Z test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_False_is_causal_True_float16_cuda_float16 SKIPPED [0.0002s] (Flash Attention was not built for this system) [ 99%] 2024-08-06T21:15:52.3079483Z test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_True_is_causal_False_bfloat16_cuda_bfloat16 SKIPPED [0.0002s] (Flash Attention was not built for this system) [ 99%] 2024-08-06T21:15:52.3080725Z test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_True_is_causal_False_float16_cuda_float16 SKIPPED [0.0002s] (Flash Attention was not built for this system) [ 99%] 2024-08-06T21:15:52.3081978Z test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_True_is_causal_True_bfloat16_cuda_bfloat16 SKIPPED [0.0002s] (Flash Attention was not built for this system) [ 99%] 2024-08-06T21:15:52.3083208Z test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_flash_attention_grad_against_math_contiguous_inputs_True_is_causal_True_float16_cuda_float16 SKIPPED [0.0002s] (Flash Attention was not built for this system) [ 99%] 2024-08-06T21:15:52.3084108Z test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_mem_efficient_grad_against_math_contiguous_inputs_False_is_causal_False_cuda PASSED [0.0037s] [ 99%] 2024-08-06T21:15:52.3084993Z test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_mem_efficient_grad_against_math_contiguous_inputs_False_is_causal_True_cuda PASSED [0.0034s] [ 99%] 2024-08-06T21:15:52.3085929Z test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_mem_efficient_grad_against_math_contiguous_inputs_True_is_causal_False_cuda PASSED [0.0033s] [ 99%] 2024-08-06T21:15:52.3086863Z test_transformers.py::TestSDPACudaOnlyCUDA::test_sdp_mem_efficient_grad_against_math_contiguous_inputs_True_is_causal_True_cuda PASSED [0.0035s] [ 99%] 2024-08-06T21:15:52.3087728Z test_transformers.py::TestSDPACudaOnlyCUDA::test_singelton_head_dim_stride_ne_1_cuda SKIPPED [0.0002s] (Fused SDPA was not built for this system) [ 99%] 2024-08-06T21:15:52.3088550Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_LOWER_RIGHT_shape0_cuda PASSED [0.0489s] [ 99%] 2024-08-06T21:15:52.3089367Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_LOWER_RIGHT_shape1_cuda PASSED [0.0990s] [ 99%] 2024-08-06T21:15:52.3090694Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_LOWER_RIGHT_shape2_cuda SKIPPED [0.0012s] (Lower right causal mask will produce NaNs in the output when seq_len_q > seq_len_kv!) [ 99%] 2024-08-06T21:15:52.3091524Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_LOWER_RIGHT_shape3_cuda PASSED [0.0041s] [ 99%] 2024-08-06T21:15:52.3092354Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_UPPER_LEFT_shape0_cuda PASSED [0.0489s] [ 99%] 2024-08-06T21:15:52.3093160Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_UPPER_LEFT_shape1_cuda PASSED [0.2474s] [ 99%] 2024-08-06T21:15:52.3093961Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_UPPER_LEFT_shape2_cuda PASSED [0.0939s] [ 99%] 2024-08-06T21:15:52.3094772Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_causal_variant_CausalVariant_UPPER_LEFT_shape3_cuda PASSED [0.0043s] [ 99%] 2024-08-06T21:15:52.3095680Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_LOWER_RIGHT_shape0_cuda PASSED [0.7785s] [ 99%] 2024-08-06T21:15:52.3096647Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_LOWER_RIGHT_shape1_cuda PASSED [0.2169s] [ 99%] 2024-08-06T21:15:52.3097952Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_LOWER_RIGHT_shape2_cuda SKIPPED [0.0031s] (Lower right causal mask will produce NaNs in the output when seq_len_q > seq_len_kv!) [ 99%] 2024-08-06T21:15:52.3098808Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_LOWER_RIGHT_shape3_cuda PASSED [0.1632s] [ 99%] 2024-08-06T21:15:52.3099672Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_UPPER_LEFT_shape0_cuda PASSED [0.1257s] [ 99%] 2024-08-06T21:15:52.3100530Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_UPPER_LEFT_shape1_cuda PASSED [0.1515s] [ 99%] 2024-08-06T21:15:52.3101392Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_UPPER_LEFT_shape2_cuda PASSED [0.1735s] [ 99%] 2024-08-06T21:15:52.3102251Z test_transformers.py::TestAttnBiasCUDA::test_causal_variants_compile_causal_variant_CausalVariant_UPPER_LEFT_shape3_cuda PASSED [0.0865s] [ 99%] 2024-08-06T21:15:52.3102843Z test_transformers.py::TestAttnBiasCUDA::test_is_causal_and_mask_fails_cuda PASSED [0.0021s] [ 99%] 2024-08-06T21:15:52.3103500Z test_transformers.py::TestAttnBiasCUDA::test_is_causal_equals_upper_left_shape0_cuda PASSED [0.0109s] [ 99%] 2024-08-06T21:15:52.3104650Z test_transformers.py::TestAttnBiasCUDA::test_is_causal_equals_upper_left_shape1_cuda PASSED [0.0117s] [ 99%] 2024-08-06T21:15:52.3105433Z test_transformers.py::TestAttnBiasCUDA::test_is_causal_equals_upper_left_shape2_cuda PASSED [0.0230s] [ 99%] 2024-08-06T21:15:52.3106082Z test_transformers.py::TestAttnBiasCUDA::test_is_causal_equals_upper_left_shape3_cuda PASSED [0.0023s] [100%] 2024-08-06T21:15:52.3106105Z 2024-08-06T21:15:52.3106778Z - generated xml file: /var/lib/jenkins/workspace/test/test-reports/python-pytest/test_transformers/test_transformers-00a63c3f2f793df3.xml - 2024-08-06T21:15:52.3107373Z ======== 7729 passed, 11 skipped, 37604 deselected in 120.67s (0:02:00) ======== 2024-08-06T21:15:52.3108591Z The following tests failed and then succeeded when run in a new process['test/test_transformers.py::TestSDPACudaOnlyCUDA::test_mem_efficient_attention_attn_mask_vs_math_ref_grads_batch_size_1_seq_len_q_128_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_cuda_float16'] 2024-08-06T21:15:52.3108613Z 2024-08-06T21:15:52.3109095Z FINISHED PRINTING LOG FILE of test_transformers 1/1 (test/test-reports/test_transformers_1.1_95af2e2cb07b8232_.log) 2024-08-06T21:15:52.3109116Z 2024-08-06T21:15:52.3109338Z Running test_proxy_tensor 1/1 ... [2024-08-06 21:15:40.444470] 2024-08-06T21:15:52.3110348Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_proxy_tensor.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:15:40.444973] 2024-08-06T21:16:02.8508248Z 2024-08-06T21:16:02.8509663Z test_proxy_tensor 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_proxy_tensor_1.1_c073645f88a0b17f_.log 2024-08-06T21:16:02.8576282Z Running 172 items in this shard: test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_allclose, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_amp_cache, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_constant_blowup, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_constant_proxy_tensor_mut, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_constant_random, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_constant_unbind, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_decomp_of_capture, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_decomposition_interpreter, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_empty_like_doesnt_burn_in_defaults, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_inplace_metadata, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_isolated_graphmodule, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_make_fx_model_double_param, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_make_fx_model_fwd_bwd, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_make_fx_model_fwd_bwd_wgtupdate, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_make_fx_overloads, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_make_fx_reentrant_dispatch, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_make_fx_simple, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_mode_tracing_factory_function, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_partial_decomp, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_pickle_issue89626, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_pr_86917, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_pre_dispatch_functionalization, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_pre_dispatch_functionalization_view_op, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_pre_dispatch_linear, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_pre_dispatch_mode_stack, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_pre_dispatch_no_grad, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_proxy_tensor, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_proxy_tensor_mode_with_decomp_table_preserves_proxy, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_resnet18_backward_trace, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_scalar_device, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_strides, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_tensor_constants, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_trace_subclasses, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_val_metadata_mutation, test/test_proxy_tensor.py::TestGenericProxyTensorReal::test_varargs, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_allclose, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_amp_cache, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_constant_blowup, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_constant_proxy_tensor_mut, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_constant_random, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_constant_unbind, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_decomp_of_capture, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_decomposition_interpreter, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_empty_like_doesnt_burn_in_defaults, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_inplace_metadata, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_isolated_graphmodule, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_make_fx_model_double_param, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_make_fx_model_fwd_bwd, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_make_fx_model_fwd_bwd_wgtupdate, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_make_fx_overloads, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_make_fx_reentrant_dispatch, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_make_fx_simple, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_mode_tracing_factory_function, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_partial_decomp, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_pickle_issue89626, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_pr_86917, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_pre_dispatch_functionalization, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_pre_dispatch_functionalization_view_op, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_pre_dispatch_linear, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_pre_dispatch_mode_stack, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_pre_dispatch_no_grad, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_proxy_tensor, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_proxy_tensor_mode_with_decomp_table_preserves_proxy, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_resnet18_backward_trace, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_scalar_device, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_strides, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_tensor_constants, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_trace_subclasses, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_val_metadata_mutation, test/test_proxy_tensor.py::TestGenericProxyTensorFake::test_varargs, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_allclose, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_amp_cache, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_constant_blowup, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_constant_proxy_tensor_mut, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_constant_random, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_constant_unbind, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_decomp_of_capture, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_decomposition_interpreter, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_empty_like_doesnt_burn_in_defaults, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_inplace_metadata, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_isolated_graphmodule, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_make_fx_model_double_param, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_make_fx_model_fwd_bwd, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_make_fx_model_fwd_bwd_wgtupdate, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_make_fx_overloads, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_make_fx_reentrant_dispatch, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_make_fx_simple, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_mode_tracing_factory_function, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_partial_decomp, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_pickle_issue89626, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_pr_86917, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_pre_dispatch_functionalization, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_pre_dispatch_functionalization_view_op, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_pre_dispatch_linear, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_pre_dispatch_mode_stack, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_pre_dispatch_no_grad, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_proxy_tensor, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_proxy_tensor_mode_with_decomp_table_preserves_proxy, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_resnet18_backward_trace, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_scalar_device, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_strides, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_tensor_constants, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_trace_subclasses, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_val_metadata_mutation, test/test_proxy_tensor.py::TestGenericProxyTensorSymbolic::test_varargs, test/test_proxy_tensor.py::TestRealProxyTensor::test_error_on_data_dependent_ops, test/test_proxy_tensor.py::TestFakeProxyTensor::test_alias, test/test_proxy_tensor.py::TestFakeProxyTensor::test_free_fake, test/test_proxy_tensor.py::TestFakeProxyTensor::test_fused_adam, test/test_proxy_tensor.py::TestFakeProxyTensor::test_issue82547, test/test_proxy_tensor.py::TestFakeProxyTensor::test_meta, test/test_proxy_tensor.py::TestFakeProxyTensor::test_use_fake_and_tensor, test/test_proxy_tensor.py::TestSymbolicTracing::test_adv_index_batch, test/test_proxy_tensor.py::TestSymbolicTracing::test_arange_unbacked_output_size, test/test_proxy_tensor.py::TestSymbolicTracing::test_binary_broadcast, test/test_proxy_tensor.py::TestSymbolicTracing::test_boolean_index, test/test_proxy_tensor.py::TestSymbolicTracing::test_broadcast_shapes, test/test_proxy_tensor.py::TestSymbolicTracing::test_cat, test/test_proxy_tensor.py::TestSymbolicTracing::test_constant_specialization, test/test_proxy_tensor.py::TestSymbolicTracing::test_cpu_scalar_cuda, test/test_proxy_tensor.py::TestSymbolicTracing::test_cumsum_unbacked, test/test_proxy_tensor.py::TestSymbolicTracing::test_debug_interpreter, test/test_proxy_tensor.py::TestSymbolicTracing::test_deduped_shape, test/test_proxy_tensor.py::TestSymbolicTracing::test_dynamic_pointwise_scalar, test/test_proxy_tensor.py::TestSymbolicTracing::test_elementwise_meta_with_sym_numbers, test/test_proxy_tensor.py::TestSymbolicTracing::test_expand, test/test_proxy_tensor.py::TestSymbolicTracing::test_fake_tensor_as_size, test/test_proxy_tensor.py::TestSymbolicTracing::test_guard_lowerbound_range_refinement, test/test_proxy_tensor.py::TestSymbolicTracing::test_guard_lowerbound_range_refinement_multivariate, test/test_proxy_tensor.py::TestSymbolicTracing::test_guard_upperbound_range_refinement, test/test_proxy_tensor.py::TestSymbolicTracing::test_guard_upperbound_range_refinement_multivariate, test/test_proxy_tensor.py::TestSymbolicTracing::test_guards_equal, test/test_proxy_tensor.py::TestSymbolicTracing::test_int_input, test/test_proxy_tensor.py::TestSymbolicTracing::test_invalidate_nonzero, test/test_proxy_tensor.py::TestSymbolicTracing::test_invalidate_nonzero_propagate_real_tensors, test/test_proxy_tensor.py::TestSymbolicTracing::test_item, test/test_proxy_tensor.py::TestSymbolicTracing::test_item_to_constructor, test/test_proxy_tensor.py::TestSymbolicTracing::test_make_fx_with_custom_tracer_preserving_nn_module_stack, test/test_proxy_tensor.py::TestSymbolicTracing::test_mega_guard, test/test_proxy_tensor.py::TestSymbolicTracing::test_metadata, test/test_proxy_tensor.py::TestSymbolicTracing::test_metadata_fresh, test/test_proxy_tensor.py::TestSymbolicTracing::test_mod_gcd_unbacked, test/test_proxy_tensor.py::TestSymbolicTracing::test_multiply_shape, test/test_proxy_tensor.py::TestSymbolicTracing::test_neg_shape, test/test_proxy_tensor.py::TestSymbolicTracing::test_new_empty, test/test_proxy_tensor.py::TestSymbolicTracing::test_non_deduped_shape, test/test_proxy_tensor.py::TestSymbolicTracing::test_non_symint_size_spec, test/test_proxy_tensor.py::TestSymbolicTracing::test_nonidentity_transitive_guards, test/test_proxy_tensor.py::TestSymbolicTracing::test_reflect_r_over_x, test/test_proxy_tensor.py::TestSymbolicTracing::test_repeat_interleave, test/test_proxy_tensor.py::TestSymbolicTracing::test_repeat_interleave_unbacked_output_size, test/test_proxy_tensor.py::TestSymbolicTracing::test_reshape_divisibility_unbacked, test/test_proxy_tensor.py::TestSymbolicTracing::test_resize_from_zero, test/test_proxy_tensor.py::TestSymbolicTracing::test_return_symint, test/test_proxy_tensor.py::TestSymbolicTracing::test_rmethod, test/test_proxy_tensor.py::TestSymbolicTracing::test_setitem_symint, test/test_proxy_tensor.py::TestSymbolicTracing::test_size_with_tensor, test/test_proxy_tensor.py::TestSymbolicTracing::test_split_unbacked_sizes, test/test_proxy_tensor.py::TestSymbolicTracing::test_sqrt_size, test/test_proxy_tensor.py::TestSymbolicTracing::test_sym_storage_offset, test/test_proxy_tensor.py::TestSymbolicTracing::test_symbolic_repeat_interleave, test/test_proxy_tensor.py::TestSymbolicTracing::test_symint_to_tensor, test/test_proxy_tensor.py::TestSymbolicTracing::test_tensor_symfloat, test/test_proxy_tensor.py::TestSymbolicTracing::test_unary, test/test_proxy_tensor.py::TestSymbolicTracing::test_unbacked_batch_resnet, test/test_proxy_tensor.py::TestSymbolicTracing::test_unbacked_slice, test/test_proxy_tensor.py::TestSymbolicTracing::test_unbacked_unification, test/test_proxy_tensor.py::TestSymbolicTracing::test_unbacked_unify_dependency_violation, test/test_proxy_tensor.py::TestSymbolicTracing::test_unbacked_unify_guard, test/test_proxy_tensor.py::TestSymbolicTracing::test_unbacked_unify_guard_transitivity, test/test_proxy_tensor.py::TestSymbolicTracing::test_view_divisibility_unbacked, test/test_proxy_tensor.py::TestSymbolicTracing::test_view_divisibility_unbacked_relatively_prime 2024-08-06T21:16:02.8642593Z 2024-08-06T21:16:06.5369142Z Running test_foreach 1/1 ... [2024-08-06 21:16:06.536427] 2024-08-06T21:16:06.5374045Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_foreach.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:16:06.536994] 2024-08-06T21:25:45.4333237Z 2024-08-06T21:25:45.4336412Z test_foreach 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_foreach_1.1_da04e0ec03200e2d_.log 2024-08-06T21:25:45.5955785Z Running 3463 items in this shard: test/test_foreach.py::TestForeachCUDA::test_0dim_tensor_overload_cpu_ok_cuda, test/test_foreach.py::TestForeachCUDA::test_0dim_tensor_overload_exception_cuda, test/test_foreach.py::TestForeachCUDA::test_add_scalar_with_empty_list_and_empty_tensor_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_add_scalar_with_empty_list_and_empty_tensor_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_add_scalar_with_empty_list_and_empty_tensor_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_add_scalar_with_empty_list_and_empty_tensor_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_add_scalar_with_empty_list_and_empty_tensor_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_add_scalar_with_empty_list_and_empty_tensor_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_add_scalar_with_empty_list_and_empty_tensor_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_add_scalar_with_empty_list_and_empty_tensor_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_add_scalar_with_empty_list_and_empty_tensor_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_add_scalar_with_empty_list_and_empty_tensor_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_add_scalar_with_empty_list_and_empty_tensor_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_abs_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_acos_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_add_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_addcdiv_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_addcmul_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_asin_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_atan_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_ceil_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_clamp_max_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_clamp_min_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_copy_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_cos_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_cosh_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_div_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_erf_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_erfc_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_exp_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_expm1_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_floor_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_frac_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_lerp_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_lgamma_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_log10_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_log1p_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_log2_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_log_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_max_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_maximum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_minimum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_mul_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_neg_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_norm_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_pow_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_reciprocal_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_round_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_sigmoid_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_sign_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_sin_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_sinh_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_sqrt_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_sub_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_tan_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_tanh_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_trunc_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_all_zero_size_tensors_do_not_launch_kernel__foreach_zero_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_abs_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_abs_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_abs_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_abs_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_acos_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_acos_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_acos_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_acos_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_add_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_add_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_add_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_add_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_addcdiv_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_addcdiv_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_addcdiv_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_addcdiv_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_addcmul_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_addcmul_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_addcmul_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_addcmul_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_asin_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_asin_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_asin_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_asin_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_atan_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_atan_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_atan_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_atan_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_ceil_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_ceil_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_ceil_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_ceil_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_clamp_max_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_clamp_max_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_clamp_max_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_clamp_max_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_clamp_min_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_clamp_min_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_clamp_min_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_clamp_min_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_copy_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_copy_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_copy_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_copy_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_cos_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_cos_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_cos_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_cos_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_cosh_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_cosh_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_cosh_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_cosh_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_div_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_div_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_div_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_div_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_erf_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_erf_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_erf_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_erf_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_erfc_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_erfc_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_erfc_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_erfc_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_exp_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_exp_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_exp_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_exp_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_expm1_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_expm1_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_expm1_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_expm1_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_floor_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_floor_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_floor_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_floor_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_frac_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_frac_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_frac_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_frac_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_lerp_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_lerp_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_lerp_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_lerp_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_lgamma_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_lgamma_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_lgamma_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_lgamma_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log10_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log10_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log10_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log10_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log1p_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log1p_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log1p_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log1p_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log2_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log2_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log2_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log2_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_log_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_max_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_max_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_max_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_max_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_maximum_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_maximum_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_maximum_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_maximum_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_minimum_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_minimum_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_minimum_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_minimum_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_mul_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_mul_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_mul_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_mul_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_neg_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_neg_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_neg_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_neg_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_norm_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_norm_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_norm_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_norm_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_pow_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_pow_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_pow_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_pow_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_reciprocal_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_reciprocal_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_reciprocal_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_reciprocal_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_round_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_round_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_round_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_round_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sigmoid_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sigmoid_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sigmoid_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sigmoid_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sign_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sign_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sign_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sign_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sin_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sin_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sin_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sin_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sinh_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sinh_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sinh_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sinh_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sqrt_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sqrt_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sqrt_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sqrt_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sub_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sub_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sub_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_sub_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_tan_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_tan_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_tan_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_tan_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_tanh_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_tanh_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_tanh_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_tanh_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_trunc_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_trunc_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_trunc_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_trunc_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_zero_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_zero_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_zero_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_autodiff__foreach_zero_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_big_num_tensors__foreach_max_use_cuda_graph_False_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_big_num_tensors__foreach_max_use_cuda_graph_False_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_big_num_tensors__foreach_max_use_cuda_graph_True_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_big_num_tensors__foreach_max_use_cuda_graph_True_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_big_num_tensors__foreach_norm_use_cuda_graph_False_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_big_num_tensors__foreach_norm_use_cuda_graph_False_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_big_num_tensors__foreach_norm_use_cuda_graph_True_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_big_num_tensors__foreach_norm_use_cuda_graph_True_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_add_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_add_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_add_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_add_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_clamp_max_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_clamp_max_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_clamp_max_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_clamp_max_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_clamp_min_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_clamp_min_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_clamp_min_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_clamp_min_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_div_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_div_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_div_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_div_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_maximum_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_maximum_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_maximum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_maximum_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_minimum_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_minimum_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_minimum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_minimum_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_mul_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_mul_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_mul_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_mul_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_pow_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_pow_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_pow_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_pow_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_sub_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_sub_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_sub_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_float_inf_nan__foreach_sub_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_add_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_add_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_add_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_add_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_add_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_add_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_add_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_add_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_add_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_add_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_add_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_add_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_max_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_max_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_max_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_max_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_max_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_max_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_max_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_max_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_max_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_max_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_max_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_max_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_min_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_min_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_min_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_min_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_min_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_min_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_min_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_min_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_min_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_min_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_min_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_clamp_min_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_div_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_div_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_div_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_div_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_div_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_div_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_div_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_div_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_div_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_div_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_div_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_div_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_maximum_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_maximum_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_maximum_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_maximum_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_maximum_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_maximum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_maximum_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_maximum_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_maximum_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_maximum_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_maximum_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_maximum_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_minimum_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_minimum_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_minimum_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_minimum_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_minimum_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_minimum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_minimum_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_minimum_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_minimum_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_minimum_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_minimum_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_minimum_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_mul_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_mul_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_mul_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_mul_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_mul_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_mul_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_mul_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_mul_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_mul_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_mul_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_mul_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_mul_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_pow_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_pow_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_pow_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_pow_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_pow_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_pow_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_pow_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_pow_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_pow_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_pow_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_pow_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_pow_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_sub_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_sub_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_sub_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_sub_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_sub_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_sub_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_sub_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_sub_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_sub_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_sub_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_sub_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_error_cases__foreach_sub_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_add_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_add_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_add_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_add_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_add_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_add_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_add_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_add_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_add_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_add_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_add_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_add_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_max_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_max_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_max_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_max_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_max_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_max_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_max_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_max_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_max_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_max_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_max_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_max_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_min_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_min_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_min_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_min_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_min_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_min_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_min_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_min_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_min_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_min_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_min_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_clamp_min_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_div_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_div_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_div_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_div_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_div_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_div_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_div_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_div_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_div_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_div_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_div_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_div_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_maximum_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_maximum_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_maximum_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_maximum_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_maximum_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_maximum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_maximum_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_maximum_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_maximum_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_maximum_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_maximum_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_maximum_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_minimum_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_minimum_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_minimum_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_minimum_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_minimum_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_minimum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_minimum_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_minimum_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_minimum_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_minimum_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_minimum_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_minimum_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_mul_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_mul_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_mul_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_mul_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_mul_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_mul_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_mul_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_mul_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_mul_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_mul_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_mul_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_mul_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_pow_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_pow_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_pow_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_pow_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_pow_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_pow_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_pow_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_pow_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_pow_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_pow_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_pow_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_pow_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_sub_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_sub_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_sub_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_sub_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_sub_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_sub_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_sub_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_sub_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_sub_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_sub_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_sub_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_list_slow_path__foreach_sub_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_different_tensor_dtypes__foreach_add_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_different_tensor_dtypes__foreach_clamp_max_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_different_tensor_dtypes__foreach_clamp_min_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_different_tensor_dtypes__foreach_div_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_different_tensor_dtypes__foreach_maximum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_different_tensor_dtypes__foreach_minimum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_different_tensor_dtypes__foreach_mul_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_different_tensor_dtypes__foreach_pow_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_different_tensor_dtypes__foreach_sub_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_add_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_add_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_add_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_add_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_add_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_add_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_add_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_add_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_add_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_add_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_add_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_add_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_max_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_max_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_max_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_max_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_max_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_max_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_max_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_max_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_max_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_max_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_max_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_max_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_min_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_min_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_min_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_min_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_min_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_min_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_min_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_min_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_min_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_min_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_min_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_clamp_min_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_div_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_div_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_div_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_div_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_div_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_div_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_div_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_div_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_div_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_div_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_div_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_div_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_maximum_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_maximum_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_maximum_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_maximum_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_maximum_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_maximum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_maximum_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_maximum_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_maximum_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_maximum_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_maximum_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_maximum_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_minimum_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_minimum_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_minimum_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_minimum_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_minimum_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_minimum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_minimum_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_minimum_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_minimum_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_minimum_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_minimum_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_minimum_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_mul_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_mul_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_mul_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_mul_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_mul_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_mul_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_mul_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_mul_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_mul_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_mul_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_mul_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_mul_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_pow_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_pow_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_pow_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_pow_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_pow_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_pow_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_pow_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_pow_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_pow_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_pow_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_pow_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_pow_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_sub_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_sub_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_sub_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_sub_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_sub_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_sub_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_sub_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_sub_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_sub_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_sub_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_sub_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_scalar_with_overlapping_tensors__foreach_sub_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_add_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_add_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_add_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_add_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_add_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_add_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_add_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_add_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_add_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_add_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_add_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_add_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_max_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_max_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_max_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_max_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_max_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_max_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_max_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_max_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_max_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_max_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_max_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_max_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_min_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_min_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_min_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_min_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_min_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_min_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_min_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_min_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_min_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_min_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_min_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_clamp_min_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_div_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_div_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_div_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_div_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_div_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_div_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_div_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_div_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_div_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_div_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_div_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_div_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_maximum_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_maximum_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_maximum_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_maximum_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_maximum_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_maximum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_maximum_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_maximum_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_maximum_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_maximum_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_maximum_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_maximum_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_minimum_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_minimum_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_minimum_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_minimum_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_minimum_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_minimum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_minimum_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_minimum_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_minimum_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_minimum_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_minimum_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_minimum_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_mul_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_mul_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_mul_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_mul_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_mul_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_mul_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_mul_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_mul_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_mul_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_mul_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_mul_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_mul_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_pow_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_pow_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_pow_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_pow_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_pow_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_pow_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_pow_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_pow_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_pow_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_pow_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_pow_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_pow_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_sub_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_sub_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_sub_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_sub_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_sub_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_sub_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_sub_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_sub_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_sub_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_sub_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_sub_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_tensors_on_different_devices__foreach_sub_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_False_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_False_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_False_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_False_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_False_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_False_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_False_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_False_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_False_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_False_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_False_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_False_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_div_reciprocal_cuda, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_device_inputs__foreach_copy_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_device_inputs__foreach_copy_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_device_inputs__foreach_copy_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_device_inputs__foreach_copy_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_device_inputs__foreach_copy_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_device_inputs__foreach_copy_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_device_inputs__foreach_copy_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_device_inputs__foreach_copy_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_device_inputs__foreach_copy_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_device_inputs__foreach_copy_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_device_inputs__foreach_copy_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_device_inputs__foreach_copy_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_foreach_l2_large_value_input__foreach_norm_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_foreach_l2_large_value_input__foreach_norm_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_max_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_max_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_max_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_max_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_max_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_max_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_max_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_max_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_max_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_max_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_max_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_max_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_norm_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_norm_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_norm_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_norm_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_norm_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_norm_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_norm_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_norm_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_norm_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_norm_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_norm_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_foreach_reduce_large_input__foreach_norm_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_abs_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_acos_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_add_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_addcdiv_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_addcmul_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_asin_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_atan_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_ceil_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_clamp_max_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_clamp_min_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_copy_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_cos_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_cosh_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_div_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_erf_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_erfc_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_exp_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_expm1_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_floor_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_frac_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_lerp_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_lgamma_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_log10_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_log1p_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_log2_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_log_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_maximum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_minimum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_mul_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_neg_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_pow_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_reciprocal_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_round_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_sigmoid_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_sign_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_sin_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_sinh_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_sqrt_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_sub_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_tan_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_tanh_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_trunc_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_inplace_foreach_leaf_check_and_grad_fn__foreach_zero_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_lifetime_of_grad_fn_when_result_is_saved__foreach_exp_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_lifetime_of_grad_fn_when_result_is_saved__foreach_expm1_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_lifetime_of_grad_fn_when_result_is_saved__foreach_pow_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_lifetime_of_grad_fn_when_result_is_saved__foreach_reciprocal_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_lifetime_of_grad_fn_when_result_is_saved__foreach_sigmoid_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_lifetime_of_grad_fn_when_result_is_saved__foreach_sqrt_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_lifetime_of_grad_fn_when_result_is_saved__foreach_tan_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_lifetime_of_grad_fn_when_result_is_saved__foreach_tanh_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_abs_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_acos_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_add_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_addcdiv_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_addcmul_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_asin_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_atan_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_ceil_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_clamp_max_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_clamp_min_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_cos_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_cosh_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_div_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_erf_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_erfc_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_exp_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_expm1_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_floor_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_frac_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_lerp_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_lgamma_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_log10_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_log1p_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_log2_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_log_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_maximum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_minimum_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_mul_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_neg_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_pow_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_reciprocal_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_round_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_sigmoid_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_sign_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_sin_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_sinh_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_sqrt_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_sub_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_tan_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_tanh_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_outplace_with_invalid_grads__foreach_trunc_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_abs_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_acos_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_add_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcdiv_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_addcmul_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_asin_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_atan_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_ceil_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_max_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_clamp_min_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_copy_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cos_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_cosh_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_div_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erf_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_erfc_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_exp_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_expm1_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_floor_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_frac_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lerp_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_lgamma_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log10_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log1p_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log2_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_log_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_max_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_maximum_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_minimum_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_mul_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_neg_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_norm_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_pow_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_reciprocal_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_round_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sigmoid_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sign_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sin_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sinh_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sqrt_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_sub_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tan_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_tanh_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_trunc_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_fastpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_inplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_inplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_inplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_inplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_inplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_inplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_inplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_inplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_inplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_inplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_inplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_inplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_outplace_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_outplace_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_outplace_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_outplace_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_outplace_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_outplace_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_outplace_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_outplace_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_outplace_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_outplace_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_outplace_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_parity__foreach_zero_slowpath_outplace_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_tensors_on_different_devices__foreach_addcdiv_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_tensors_on_different_devices__foreach_addcdiv_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_tensors_on_different_devices__foreach_addcmul_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_tensors_on_different_devices__foreach_addcmul_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_False_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_False_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_False_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_False_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_False_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_False_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_False_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_False_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_False_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_False_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_False_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_False_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_True_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_True_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_True_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_True_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_True_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_True_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_True_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_True_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_True_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_True_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_True_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcdiv_is_fastpath_True_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_False_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_False_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_False_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_False_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_False_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_False_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_False_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_False_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_False_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_False_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_False_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_False_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_True_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_True_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_True_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_True_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_True_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_True_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_True_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_True_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_True_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_True_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_True_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_pointwise_op_with_tensor_of_scalarlist_overload__foreach_addcmul_is_fastpath_True_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_tensors_grouping_cuda, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_abs_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_abs_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_abs_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_abs_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_abs_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_abs_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_abs_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_abs_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_abs_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_abs_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_abs_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_abs_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_acos_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_acos_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_acos_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_acos_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_acos_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_acos_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_acos_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_acos_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_acos_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_acos_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_acos_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_acos_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_asin_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_asin_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_asin_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_asin_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_asin_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_asin_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_asin_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_asin_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_asin_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_asin_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_asin_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_asin_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_atan_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_atan_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_atan_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_atan_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_atan_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_atan_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_atan_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_atan_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_atan_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_atan_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_atan_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_atan_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_ceil_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_ceil_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_ceil_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_ceil_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_ceil_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_ceil_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_ceil_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_ceil_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_ceil_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_ceil_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_ceil_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_ceil_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cos_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cos_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cos_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cos_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cos_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cos_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cos_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cos_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cos_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cos_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cos_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cos_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cosh_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cosh_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cosh_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cosh_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cosh_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cosh_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cosh_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cosh_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cosh_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cosh_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cosh_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_cosh_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erf_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erf_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erf_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erf_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erf_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erf_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erf_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erf_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erf_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erf_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erf_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erf_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erfc_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erfc_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erfc_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erfc_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erfc_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erfc_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erfc_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erfc_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erfc_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erfc_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erfc_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_erfc_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_exp_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_exp_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_exp_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_exp_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_exp_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_exp_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_exp_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_exp_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_exp_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_exp_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_exp_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_exp_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_expm1_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_expm1_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_expm1_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_expm1_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_expm1_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_expm1_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_expm1_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_expm1_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_expm1_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_expm1_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_expm1_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_expm1_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_floor_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_floor_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_floor_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_floor_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_floor_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_floor_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_floor_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_floor_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_floor_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_floor_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_floor_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_floor_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_frac_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_frac_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_frac_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_frac_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_frac_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_frac_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_frac_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_frac_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_frac_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_frac_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_frac_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_frac_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_lgamma_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_lgamma_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_lgamma_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_lgamma_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_lgamma_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_lgamma_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_lgamma_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_lgamma_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_lgamma_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_lgamma_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_lgamma_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_lgamma_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log10_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log10_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log10_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log10_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log10_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log10_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log10_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log10_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log10_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log10_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log10_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log10_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log1p_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log1p_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log1p_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log1p_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log1p_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log1p_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log1p_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log1p_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log1p_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log1p_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log1p_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log1p_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log2_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log2_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log2_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log2_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log2_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log2_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log2_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log2_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log2_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log2_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log2_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log2_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_log_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_neg_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_neg_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_neg_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_neg_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_neg_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_neg_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_neg_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_neg_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_neg_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_neg_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_neg_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_neg_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_reciprocal_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_reciprocal_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_reciprocal_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_reciprocal_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_reciprocal_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_reciprocal_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_reciprocal_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_reciprocal_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_reciprocal_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_reciprocal_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_reciprocal_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_reciprocal_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_round_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_round_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_round_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_round_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_round_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_round_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_round_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_round_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_round_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_round_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_round_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_round_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sigmoid_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sigmoid_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sigmoid_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sigmoid_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sigmoid_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sigmoid_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sigmoid_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sigmoid_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sigmoid_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sigmoid_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sigmoid_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sigmoid_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sign_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sign_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sign_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sign_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sign_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sign_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sign_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sign_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sign_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sign_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sign_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sign_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sin_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sin_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sin_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sin_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sin_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sin_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sin_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sin_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sin_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sin_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sin_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sin_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sinh_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sinh_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sinh_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sinh_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sinh_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sinh_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sinh_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sinh_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sinh_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sinh_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sinh_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sinh_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sqrt_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sqrt_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sqrt_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sqrt_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sqrt_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sqrt_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sqrt_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sqrt_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sqrt_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sqrt_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sqrt_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_sqrt_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tan_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tan_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tan_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tan_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tan_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tan_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tan_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tan_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tan_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tan_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tan_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tan_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tanh_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tanh_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tanh_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tanh_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tanh_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tanh_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tanh_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tanh_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tanh_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tanh_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tanh_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_tanh_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_trunc_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_trunc_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_trunc_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_trunc_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_trunc_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_trunc_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_trunc_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_trunc_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_trunc_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_trunc_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_trunc_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_trunc_cuda_uint8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_zero_cuda_bfloat16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_zero_cuda_bool, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_zero_cuda_complex128, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_zero_cuda_complex64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_zero_cuda_float16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_zero_cuda_float32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_zero_cuda_float64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_zero_cuda_int16, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_zero_cuda_int32, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_zero_cuda_int64, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_zero_cuda_int8, test/test_foreach.py::TestForeachCUDA::test_unary_op_tensors_on_different_devices__foreach_zero_cuda_uint8 2024-08-06T21:25:45.7514124Z 2024-08-06T21:25:49.0721628Z Running test_ops_gradients 1/4 ... [2024-08-06 21:25:49.071705] 2024-08-06T21:25:49.0725488Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_ops_gradients.py', '-m', 'not serial', '--shard-id=1', '--num-shards=4', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:25:49.072157] 2024-08-06T21:27:12.7221529Z 2024-08-06T21:27:12.7225280Z functorch/test_ops 8/9 was successful, full logs can be found in artifacts with path test/test-reports/functorch.test_ops_8.9_32afb2878ef66931_.log 2024-08-06T21:27:12.7743390Z Running 1141 items in this shard: test/functorch/test_ops.py::TestOperatorsCUDA::test_extremal_numerics_l1_loss_cuda, test/functorch/test_ops.py::TestOperatorsCUDA::test_extremal_numerics_mse_loss_cuda, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_NumpyTakeAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad___rdiv___cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad___rsub___cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad__unsafe_masked_index_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_addmm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_allclose_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_bernoulli_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_cdist_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_cfloat_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_cumprod_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_cumsum_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_cumulative_trapezoid_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_empty_permuted_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_erfc_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_erfinv_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_exp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_fft_fft_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_fft_hfftn_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_fft_ifftshift_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_fft_ihfft_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_fft_irfft2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_fill_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_fliplr_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_float_power_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_jiterator_binary_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_jiterator_binary_return_by_ref_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_le_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_lerp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_linalg_pinv_singular_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_linalg_solve_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_linalg_solve_ex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_linalg_vector_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_log_softmax_with_dtype_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_logical_not_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_logical_or_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_long_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_masked_argmax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_masked_fill_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_masked_prod_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_maximum_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_min_reduction_with_dim_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nanmedian_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nansum_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_native_batch_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_new_empty_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_adaptive_avg_pool2d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_channel_shuffle_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_conv2d_strided_padding_dilation_no_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_conv3d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_cosine_embedding_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_dropout3d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_elu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_hardshrink_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_interpolate_bilinear_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_linear_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_margin_ranking_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_pixel_shuffle_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_rms_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_smooth_l1_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_softshrink_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nn_functional_threshold_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nonzero_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_nonzero_static_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_norm_nuc_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_polygamma_polygamma_n_2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_prod_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_rand_like_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_randint_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_randint_like_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_randn_like_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_reciprocal_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_scatter_reduce_mean_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_select_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_sin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_special_modified_bessel_i1_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_std_mean_unbiased_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_svd_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_t_copy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_take_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_grad_tan_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_NumpyMulAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp___rsub___cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_acosh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_addbmm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_addr_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_all_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_aminmax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_argmin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_as_strided_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_as_strided_scatter_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_ceil_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_column_stack_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_complex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_conj_physical_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_count_nonzero_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_diagonal_copy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_diagonal_scatter_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_div_trunc_rounding_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_double_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_empty_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_fill_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_index_reduce_amax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_index_reduce_mean_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_jiterator_4inputs_with_extra_args_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_linalg_cross_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_linalg_diagonal_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_linalg_eig_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_linalg_householder_product_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_linalg_lstsq_grad_oriented_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_linalg_solve_triangular_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_linalg_tensorinv_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_log1p_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_logcumsumexp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_logit_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_masked_cumprod_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_masked_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_masked_softmax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_maximum_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_median_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_min_reduction_no_dim_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_native_layer_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_new_empty_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_nn_functional_avg_pool3d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_nn_functional_batch_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_nn_functional_conv_transpose2d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_nn_functional_elu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_nn_functional_hardshrink_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_nn_functional_hardtanh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_nn_functional_instance_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_nn_functional_mish_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_nn_functional_nll_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_nn_functional_relu6_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_nn_functional_softmin_with_dtype_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_nn_functional_upsample_bilinear_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_ones_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_put_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_scatter_add_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_sign_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_sparse_mm_reduce_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_special_bessel_j0_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_special_bessel_j1_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_special_entr_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_special_legendre_polynomial_p_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_special_modified_bessel_k0_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_special_shifted_chebyshev_polynomial_v_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_special_shifted_chebyshev_polynomial_w_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_special_xlog1py_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_special_zeta_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_sub_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_t_copy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_transpose_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_var_mean_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_view_as_complex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvp_vsplit_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpjvpvmap_MulGenVmapAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_H_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_NumpyMulAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp___rmul___cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_as_strided_scatter_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_asinh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_broadcast_tensors_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_chalf_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_cholesky_solve_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_cumprod_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_expand_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_fft_fft2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_fft_ifftshift_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_fft_ihfft2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_fft_irfft2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_float_power_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_floor_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_full_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_full_like_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_half_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_jiterator_binary_return_by_ref_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_ldexp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_le_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_lerp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_linalg_cholesky_ex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_linalg_householder_product_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_linalg_lstsq_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_linalg_pinv_singular_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_log1p_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_lu_solve_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_masked_logsumexp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_masked_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_masked_std_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_meshgrid_list_of_tensors_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_mode_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nansum_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_native_batch_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_adaptive_avg_pool3d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_adaptive_max_pool1d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_avg_pool3d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_batch_norm_without_cudnn_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_celu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_conv2d_stride_groups_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_conv2d_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_dropout2d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_embedding_functorch_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_hardtanh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_interpolate_bilinear_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_local_response_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_margin_ranking_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_multi_margin_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_pad_circular_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_nn_functional_soft_margin_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_normal_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_normal_number_mean_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_outer_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_polar_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_real_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_renorm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_reshape_as_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_resize__cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_resolve_neg_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_round_decimals_0_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_round_decimals_neg_3_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_scatter_reduce_amax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_short_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_sign_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_sinh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_special_bessel_y0_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_special_bessel_y1_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_split_list_args_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_take_along_dim_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_transpose_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_trapezoid_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_unflatten_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_unsqueeze_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvjp_xlogy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvmap_CubeGenVmapAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvmap_NumpyCubeAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvmap_NumpySortAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvmap_ScaleGradGenVmapAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvmap_ZeroGradientsGenVmapAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvmapvmap_MulGenVmapAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvmapvmap_SelectAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvmapvmap_SortGenVmapAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_jvpvmapvmap_ZeroGradientsGenVmapAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_view_then_inplace_diagonal_grad_op_vjp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_view_then_inplace_list_return_hsplit_grad_op_jvp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_view_then_inplace_mH_grad_op_vjp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_view_then_inplace_real_grad_op_vjp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_view_then_inplace_reshape_grad_op_vjp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_view_then_inplace_resolve_neg_grad_op_jvp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_view_then_inplace_select_grad_op_vjp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_view_then_inplace_squeeze_grad_op_jvp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_view_then_inplace_transpose_grad_op_vjp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_view_then_inplace_unflatten_grad_op_jvp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_acos_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_acosh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_atanh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_block_diag_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_cdist_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_char_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_cosh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_cummin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_diagonal_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_diagonal_scatter_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_diff_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_digamma_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_div_no_rounding_mode_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_erf_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_eye_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_fft_hfftn_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_fft_ifft2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_geometric_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_gradient_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_grid_sampler_2d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_half_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_half_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_hstack_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_int_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_kthvalue_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_linalg_diagonal_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_linalg_eigh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_linalg_inv_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_linalg_ldl_factor_ex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_linalg_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_linalg_vector_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_masked_amin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nanmean_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_new_empty_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_new_full_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_new_ones_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nn_functional_batch_norm_without_cudnn_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nn_functional_bilinear_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nn_functional_binary_cross_entropy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nn_functional_cosine_similarity_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nn_functional_logsigmoid_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nn_functional_max_pool3d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nn_functional_max_unpool1d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nn_functional_multilabel_margin_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nn_functional_multilabel_soft_margin_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nn_functional_normalize_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nn_functional_pad_circular_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nn_functional_pad_replicate_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nn_functional_softsign_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_nonzero_static_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_norm_fro_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_polygamma_polygamma_n_4_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_ravel_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_resize_as__cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_round_decimals_neg_3_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_scatter_add_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_short_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_signal_windows_bartlett_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_special_bessel_j0_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_special_i0e_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_special_zeta_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_squeeze_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_squeeze_multiple_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_triangular_solve_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_unbind_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_unsqueeze_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_vsplit_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_vstack_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjp_zeros_like_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_MulGenVmapAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_NumpyExpMarkDirtyAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp___rsub___cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp__batch_norm_with_update_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_abs_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_atleast_2d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_baddbmm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_byte_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_cholesky_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_clamp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_conj_physical_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_copysign_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_cosh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_cummax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_diagflat_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_diagonal_copy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_empty_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_expand_copy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_fft_ifft_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_fft_ihfftn_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_full_like_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_grid_sampler_2d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_index_fill_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_index_reduce_mean_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_isfinite_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_isposinf_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_jiterator_2inputs_2outputs_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_linalg_cholesky_ex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_linalg_cond_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_linalg_cross_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_linalg_inv_ex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_linalg_lstsq_grad_oriented_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_linalg_lu_solve_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_linalg_matrix_rank_hermitian_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_linalg_norm_subgradients_at_zero_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_linalg_slogdet_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_linalg_solve_ex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_linspace_tensor_overload_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_log10_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_logspace_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_long_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_lt_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_masked_argmax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_masked_cumsum_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_masked_softmin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_min_binary_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_min_reduction_no_dim_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_mul_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_narrow_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_new_full_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nn_functional_avg_pool3d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nn_functional_conv2d_stride_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nn_functional_cosine_similarity_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nn_functional_ctc_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nn_functional_elu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nn_functional_embedding_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nn_functional_fractional_max_pool2d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nn_functional_gelu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nn_functional_hardsigmoid_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nn_functional_huber_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nn_functional_leaky_relu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nn_functional_max_pool2d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nn_functional_rms_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nn_functional_selu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_nonzero_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_ops_aten_index_put_functorch_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_ormqr_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_pca_lowrank_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_randn_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_renorm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_reshape_as_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_round_decimals_neg_3_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_rsqrt_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_scatter_reduce_amin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_scatter_reduce_mean_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_select_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_short_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_signal_windows_cosine_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_signal_windows_kaiser_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_slice_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_special_legendre_polynomial_p_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_special_polygamma_special_polygamma_n_0_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_special_spherical_bessel_j0_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_special_zeta_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_split_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_std_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_stft_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_svd_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_tanh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_topk_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_torch_ops_aten__efficient_attention_forward_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_unfold_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_unsqueeze_copy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_var_unbiased_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjp_xlogy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvjpvmap_NumpySortAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_NumpySortAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap__chunk_cat_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap__unsafe_masked_index_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_allclose_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_argwhere_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_asin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_bool_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_cauchy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_char_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_count_nonzero_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_cross_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_cummax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_diagflat_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_dist_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_dsplit_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_expand_copy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_fft_fft2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_floor_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_geqrf_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_half_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_heaviside_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_i0_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_igammac_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_index_reduce_prod_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_isin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_jiterator_binary_return_by_ref_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_kron_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_linalg_ldl_solve_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_linalg_lu_solve_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_linalg_matrix_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_linalg_norm_subgradients_at_zero_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_linalg_pinv_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_linalg_pinv_hermitian_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_linalg_pinv_singular_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_linalg_vander_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_log10_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_logit_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_masked_log_softmax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_masked_sum_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_min_reduction_with_dim_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_movedim_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_multinomial_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_nn_functional_avg_pool3d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_nn_functional_batch_norm_without_cudnn_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_nn_functional_conv2d_stride_no_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_nn_functional_conv2d_strided_padding_dilation_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_nn_functional_conv_transpose2d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_nn_functional_cosine_similarity_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_nn_functional_kl_div_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_nn_functional_layer_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_nn_functional_max_pool2d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_nn_functional_multilabel_soft_margin_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_nn_functional_pdist_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_nn_functional_poisson_nll_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_nn_functional_relu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_nn_functional_softmin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_normal_number_mean_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_permute_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_polygamma_polygamma_n_1_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_polygamma_polygamma_n_2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_polygamma_polygamma_n_3_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_randn_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_ravel_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_resolve_conj_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_rot90_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_round_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_scatter_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_signal_windows_nuttall_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_sin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_special_chebyshev_polynomial_u_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_special_i0e_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_special_modified_bessel_k0_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_special_shifted_chebyshev_polynomial_u_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_special_shifted_chebyshev_polynomial_v_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_special_zeta_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_std_mean_unbiased_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_svd_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_tile_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmap_unsqueeze_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vjpvmapvmap_NumpyCubeNotComposableAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_NumpyMulAutogradFunction_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad___rmul___cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad__chunk_cat_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad__native_batch_norm_legit_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad__unsafe_masked_index_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad__unsafe_masked_index_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_aminmax_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_arange_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_argwhere_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_as_strided_copy_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_asin_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_atanh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_atleast_2d_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_baddbmm_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_bool_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_broadcast_tensors_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_bucketize_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_cdist_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_cdouble_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_char_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_char_functorch_no_channels_last_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_chunk_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_clone_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_conj_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_copysign_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_cov_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_cumprod_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_diag_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_diagflat_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_diagonal_copy_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_diff_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_dist_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_div_no_rounding_mode_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_dsplit_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_empty_strided_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_equal_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_erfinv_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_expand_as_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_fft_fft2_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_fft_fft_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_fft_fftshift_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_fft_ifftshift_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_fft_irfft_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_fft_rfft2_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_flip_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_fliplr_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_flipud_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_float_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_floor_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_full_like_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_gather_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_geometric_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_gt_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_half_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_hstack_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_index_put_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_index_reduce_amax_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_isfinite_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_kron_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_ldexp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_ldexp_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_lerp_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_linalg_cholesky_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_linalg_cholesky_ex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_linalg_eigvalsh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_linalg_ldl_factor_ex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_linalg_qr_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_linalg_solve_ex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_linalg_svd_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_linalg_svdvals_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_linalg_tensorsolve_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_linalg_vector_norm_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_linspace_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_linspace_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_linspace_tensor_overload_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_log1p_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_logdet_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_logical_xor_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_logit_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_logspace_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_masked_argmin_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_masked_cumprod_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_masked_logaddexp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_masked_mean_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_masked_norm_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_masked_sum_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_max_reduction_no_dim_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_max_reduction_with_dim_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_maximum_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_median_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_meshgrid_list_of_tensors_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_meshgrid_variadic_tensors_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_min_binary_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_min_reduction_no_dim_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_mm_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_multinomial_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_multinomial_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_mv_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_narrow_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_neg_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_adaptive_avg_pool3d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_adaptive_max_pool2d_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_avg_pool3d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_avg_pool3d_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_binary_cross_entropy_with_logits_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_conv2d_no_bias_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_conv2d_strided_padding_dilation_no_bias_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_conv2d_strided_padding_dilation_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_conv2d_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_conv_transpose1d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_conv_transpose1d_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_cosine_embedding_loss_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_fractional_max_pool3d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_glu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_grid_sample_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_hardsigmoid_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_hardtanh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_huber_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_interpolate_bicubic_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_kl_div_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_max_unpool3d_grad_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_multilabel_soft_margin_loss_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_pad_constant_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_relu6_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_selu_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_triplet_margin_with_distance_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_nn_functional_unfold_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_norm_fro_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_norm_nuc_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_normal_in_place_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_ormqr_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_ormqr_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_polygamma_polygamma_n_3_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_reciprocal_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_repeat_interleave_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_rsqrt_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_scatter_reduce_amin_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_short_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_signal_windows_gaussian_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_signal_windows_general_hamming_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_signal_windows_hamming_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_softmax_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_softmax_with_dtype_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_sort_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_sparse_sampled_addmm_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_special_entr_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_special_erfcx_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_special_hermite_polynomial_he_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_special_modified_bessel_k1_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_special_scaled_modified_bessel_k1_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_square_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_std_mean_unbiased_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_stft_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_sum_to_size_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_sum_to_size_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_svd_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_tensordot_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_torch_ops_aten__efficient_attention_forward_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_torch_ops_aten__safe_softmax_default_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_trace_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_trapz_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_triu_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_trunc_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_trunc_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_unflatten_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_unsafe_split_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_unsqueeze_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_var_unbiased_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_vdot_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_view_copy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_vstack_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmap_autograd_grad_xlogy_cuda_float64, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_T_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall___getitem___functorch_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_acos_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_addbmm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_any_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_as_strided_partial_views_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_bfloat16_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_block_diag_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_bmm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_bool_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_cholesky_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_conj_physical_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_cross_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_deg2rad_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_diagonal_copy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_dsplit_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_exp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_fft_rfft2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_ge_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_grid_sampler_2d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_half_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_SelectAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_T_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule__segment_reduce_offsets_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule__softmax_backward_data_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule__unsafe_masked_index_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_addmm_decomposed_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_all_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_amin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_angle_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_argmax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_as_strided_copy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_as_strided_partial_views_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_atan2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_block_diag_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_bool_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_broadcast_tensors_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_cauchy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_cosh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_cov_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_cross_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_cumulative_trapezoid_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_diag_embed_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_double_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_fft_hfftn_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_fft_irfft2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_geometric_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_heaviside_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_histc_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_index_reduce_amin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_jiterator_4inputs_with_extra_args_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_kthvalue_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_lgamma_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_linalg_eig_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_linalg_svdvals_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_linalg_vander_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_linalg_vecdot_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_log_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_logaddexp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_logdet_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_logical_not_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_logical_or_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_logsumexp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_masked_cumprod_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_masked_logsumexp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_max_pool2d_with_indices_backward_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_meshgrid_list_of_tensors_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_meshgrid_variadic_tensors_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_mm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_mode_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_movedim_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nan_to_num_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_native_batch_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nn_functional_adaptive_avg_pool1d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nn_functional_adaptive_max_pool1d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nn_functional_celu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nn_functional_conv2d_no_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nn_functional_conv2d_stride_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nn_functional_conv2d_strided_padding_dilation_no_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nn_functional_fractional_max_pool3d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nn_functional_glu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nn_functional_local_response_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nn_functional_max_unpool1d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nn_functional_mse_loss_functorch_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nn_functional_multi_head_attention_forward_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nn_functional_normalize_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_nn_functional_unfold_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_polar_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_polygamma_polygamma_n_1_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_repeat_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_resize_as__cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_scatter_reduce_sum_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_short_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_sigmoid_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_signal_windows_exponential_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_slice_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_softmax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_softmax_with_dtype_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_special_ndtri_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_special_xlog1py_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_torch_ops_aten__efficient_attention_forward_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_trunc_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_var_mean_unbiased_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_where_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_has_batch_rule_xlogy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_index_reduce_amax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_index_reduce_mean_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_isinf_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_jiterator_binary_return_by_ref_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_linalg_eigvals_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_linalg_lstsq_grad_oriented_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_linalg_solve_ex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_linalg_tensorsolve_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_linalg_vector_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_log_softmax_with_dtype_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_long_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_lu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_masked_cumsum_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_masked_normalize_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_masked_prod_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_max_binary_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_movedim_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_multinomial_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_mv_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nanmean_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_new_ones_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nn_functional_binary_cross_entropy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nn_functional_conv2d_stride_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nn_functional_dropout2d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nn_functional_embedding_functorch_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nn_functional_grid_sample_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nn_functional_hinge_embedding_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nn_functional_huber_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nn_functional_interpolate_nearest-exact_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nn_functional_interpolate_nearest_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nn_functional_multilabel_margin_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nn_functional_relu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nn_functional_silu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nn_functional_smooth_l1_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_nn_functional_soft_margin_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_normal_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_positive_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_remainder_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_reshape_as_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_resize__cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_scalar_tensor_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_select_scatter_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_signal_windows_general_cosine_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_signbit_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_special_i1e_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_special_xlog1py_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_stack_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_svd_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_torch_ops_aten__safe_softmax_default_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_triangular_solve_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpall_where_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_NumpySortAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_ScaleGradGenVmapAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_ZeroGradientsGenVmapAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp___rmod___cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp__segment_reduce_lengths_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_abs_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_acos_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_argmax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_as_strided_copy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_asin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_block_diag_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_broadcast_to_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_char_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_column_stack_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_cumsum_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_dstack_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_erfinv_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_fft_fft_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_fft_ifftn_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_fill_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_fmax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_full_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_i0_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_int_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_isposinf_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_ldexp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_lerp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_linalg_cross_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_linalg_diagonal_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_linalg_eigvals_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_linalg_matrix_power_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_linalg_tensorinv_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_linalg_tensorsolve_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_linspace_tensor_overload_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_logit_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_lu_solve_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_masked_amax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_masked_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_min_reduction_no_dim_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_mul_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_multinomial_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_mvlgamma_mvlgamma_p_5_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_ne_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_nn_functional_conv2d_stride_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_nn_functional_dropout_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_nn_functional_max_unpool1d_grad_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_nn_functional_normalize_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_nn_functional_rms_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_nn_functional_rrelu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_nn_functional_selu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_nn_functional_unfold_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_pinverse_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_ravel_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_repeat_interleave_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_resolve_neg_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_rot90_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_select_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_short_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_sign_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_signal_windows_gaussian_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_special_chebyshev_polynomial_w_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_special_hermite_polynomial_h_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_special_i1_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_special_laguerre_polynomial_l_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_special_modified_bessel_k0_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_special_modified_bessel_k1_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_squeeze_multiple_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_tile_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_to_sparse_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_unbind_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_unsqueeze_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvjp_zeros_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvmap_NumpyExpMarkDirtyAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvmap_NumpySortAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapjvpvmap_SelectAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_NumpyCubeNotComposableAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_SelectAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_ZeroGradientsGenVmapAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp___getitem___functorch_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp__segment_reduce_lengths_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_addbmm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_addcdiv_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_any_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_byte_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_cfloat_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_char_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_clamp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_cumulative_trapezoid_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_diagonal_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_dist_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_erf_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_fft_ihfft2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_fft_rfft_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_fill_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_floor_divide_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_SortGenVmapAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_T_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_ZeroGradientsGenVmapAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule___rpow___cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule__batch_norm_with_update_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule__chunk_cat_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule__native_batch_norm_legit_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_abs_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_argwhere_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_broadcast_shapes_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_byte_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_cdist_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_char_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_clamp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_clone_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_complex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_conj_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_cross_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_div_no_rounding_mode_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_dot_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_exp2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_fft_ifftshift_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_fft_ihfft2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_fft_rfftn_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_flipud_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_float_power_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_geqrf_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_inner_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_int_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_isfinite_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_item_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_jiterator_4inputs_with_extra_args_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_linalg_det_singular_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_linalg_eig_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_linalg_ldl_factor_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_linalg_lu_solve_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_linalg_multi_dot_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_linalg_pinv_hermitian_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_linalg_slogdet_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_linalg_svdvals_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_logical_and_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_logical_not_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_long_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_masked_fill_functorch_Scalar_only_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_masked_softmin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_ne_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_new_full_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_new_zeros_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_adaptive_max_pool2d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_batch_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_conv2d_stride_depthwise_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_conv2d_stride_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_conv2d_strided_padding_dilation_no_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_elu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_embedding_bag_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_glu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_group_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_interpolate_bilinear_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_kl_div_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_local_response_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_mish_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_normalize_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_pairwise_distance_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_pixel_shuffle_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_poisson_nll_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_prelu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_relu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_nn_functional_softmin_with_dtype_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_normal_in_place_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_polygamma_polygamma_n_3_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_pow_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_put_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_scatter_add_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_searchsorted_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_sgn_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_short_functorch_no_channels_last_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_special_bessel_j0_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_special_chebyshev_polynomial_t_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_special_entr_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_special_xlog1py_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_tensordot_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_trace_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_unique_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_has_batch_rule_unsafe_chunk_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_igammac_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_isposinf_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_linalg_inv_ex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_linalg_lstsq_grad_oriented_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_linalg_lu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_linalg_norm_subgradients_at_zero_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_linalg_svd_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_log_softmax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_logaddexp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_logspace_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_long_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_masked_logsumexp_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_masked_mean_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_masked_median_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_masked_var_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_mode_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_nanquantile_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_nn_functional_conv2d_stride_depthwise_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_nn_functional_cosine_embedding_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_nn_functional_cosine_similarity_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_nn_functional_cross_entropy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_nn_functional_gaussian_nll_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_nn_functional_gelu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_nn_functional_hardtanh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_nn_functional_interpolate_linear_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_nn_functional_max_pool1d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_nn_functional_pad_reflect_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_nn_functional_pairwise_distance_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_nn_functional_poisson_nll_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_nn_functional_tanhshrink_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_normal_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_polygamma_polygamma_n_3_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_quantile_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_rad2deg_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_reshape_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_resize__cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_scatter_reduce_amin_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_softmax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_sort_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_special_airy_ai_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_special_chebyshev_polynomial_t_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_special_legendre_polynomial_p_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_special_modified_bessel_i1_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_special_shifted_chebyshev_polynomial_v_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_split_list_args_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_squeeze_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_to_sparse_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_torch_ops_aten__safe_softmax_default_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_trace_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_unfold_copy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjp_unique_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_CubeGenVmapAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_NumpyMulAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp___getitem___cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp___rmatmul___cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_argwhere_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_atleast_1d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_bmm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_bool_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_broadcast_shapes_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_cat_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_ceil_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_cholesky_inverse_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_cosh_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_cov_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_cross_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_cumprod_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_cumulative_trapezoid_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_deg2rad_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_diagonal_scatter_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_dist_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_div_no_rounding_mode_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_empty_like_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_erfinv_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_fft_fftshift_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_fft_ifft2_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_fft_ihfftn_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_floor_divide_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_frac_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_half_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_index_select_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_isinf_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_kthvalue_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_linalg_ldl_factor_ex_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_linalg_matrix_rank_hermitian_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_linalg_vander_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_linspace_tensor_overload_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_logical_or_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_masked_fill_functorch_Scalar_only_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_masked_median_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_masked_scatter_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_masked_std_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_meshgrid_variadic_tensors_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_min_reduction_no_dim_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_msort_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_mul_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_native_batch_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_adaptive_max_pool1d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_batch_norm_without_cudnn_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_celu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_conv2d_stride_depthwise_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_conv2d_stride_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_conv2d_with_bias_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_cross_entropy_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_fractional_max_pool2d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_fractional_max_pool3d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_group_norm_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_max_pool1d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_max_unpool1d_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_nll_loss_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_pad_circular_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nn_functional_selu_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_nonzero_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_norm_fro_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_normal_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_ormqr_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_polygamma_polygamma_n_3_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_polygamma_polygamma_n_4_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_prod_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_randint_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_rot90_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_round_decimals_neg_3_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_scatter_reduce_amax_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_sigmoid_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_signbit_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_special_bessel_j0_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_special_entr_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_sub_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_tan_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_tile_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_torch_ops_aten__efficient_attention_forward_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_unsqueeze_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvjp_vdot_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvjpvmap_NumpyTakeAutogradFunction_cuda_float32, test/functorch/test_ops.py::TestOperatorsCUDA::test_vmapvmapjvp_linalg_solve_cuda 2024-08-06T21:27:12.8243943Z 2024-08-06T21:27:16.4140185Z Running test_meta 5/5 ... [2024-08-06 21:27:16.413454] 2024-08-06T21:27:16.4144407Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_meta.py', '-m', 'not serial', '--shard-id=5', '--num-shards=5', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:27:16.413954] 2024-08-06T21:33:38.2354565Z 2024-08-06T21:33:38.2355806Z test_ops_gradients 1/4 was successful, full logs can be found in artifacts with path test/test-reports/test_ops_gradients_1.4_9a2dbcbd5d685f9d_.log 2024-08-06T21:33:38.2987741Z Running 1354 items in this shard: test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_NumpyCatCustomOp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_NumpyMulCustomOp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_NumpySplitCopyWithIntCustomOp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_T_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad___rmatmul___cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad___rmatmul___cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad___rmul___cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad___rpow___cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad__batch_norm_with_update_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad__native_batch_norm_legit_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_abs_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_acos_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_addbmm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_addmm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_addr_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_angle_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_angle_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_argmax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_argsort_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_as_strided_partial_views_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_as_strided_partial_views_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_asinh_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_atan_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_atleast_2d_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_atleast_3d_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_baddbmm_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_bmm_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_bool_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_broadcast_tensors_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_broadcast_tensors_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_cat_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_cauchy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_cdouble_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_cfloat_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_chalf_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_cholesky_solve_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_chunk_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_clamp_min_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_column_stack_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_conj_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_contiguous_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_corrcoef_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_corrcoef_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_cos_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_cosh_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_cov_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_cross_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_cumsum_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_deg2rad_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_diag_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_diag_embed_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_diagonal_copy_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_diagonal_scatter_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_div_trunc_rounding_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_dot_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_double_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_double_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_einsum_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_empty_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_eq_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_erfc_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_erfinv_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_expand_as_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_expand_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_fft_hfft_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_fft_ifftn_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_fft_ihfft_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_fill_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_flatten_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_flip_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_flipud_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_float_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_float_power_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_float_power_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_full_like_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_ge_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_geqrf_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_gradient_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_grid_sampler_2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_gt_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_half_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_i0_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_index_put_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_index_reduce_mean_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_isclose_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_isinf_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_isinf_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_jiterator_4inputs_with_extra_args_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_jiterator_binary_return_by_ref_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_jiterator_binary_return_by_ref_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_kron_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_lerp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_cholesky_ex_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_cross_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_det_singular_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_det_singular_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_diagonal_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_eigh_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_eigvals_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_eigvalsh_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_inv_ex_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_ldl_factor_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_ldl_factor_ex_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_lu_factor_ex_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_lu_factor_ex_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_lu_solve_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_matrix_power_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_matrix_rank_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_multi_dot_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_norm_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_solve_ex_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_vander_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linalg_vecdot_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linspace_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linspace_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_linspace_tensor_overload_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_log_normal_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_log_softmax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_log_softmax_with_dtype_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_logcumsumexp_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_logdet_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_logical_not_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_logical_xor_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_logspace_tensor_overload_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_long_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_lu_solve_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_lu_unpack_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_mH_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_mH_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_mT_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_masked_cumsum_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_masked_cumsum_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_masked_fill_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_masked_logaddexp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_masked_mean_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_masked_scatter_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_masked_std_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_masked_sum_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_masked_var_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_matmul_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_matmul_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_matrix_exp_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_max_pool2d_with_indices_backward_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_meshgrid_list_of_tensors_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_meshgrid_variadic_tensors_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_min_binary_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_min_reduction_no_dim_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_msort_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_multinomial_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_mvlgamma_mvlgamma_p_3_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nanmean_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_narrow_copy_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_narrow_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_new_full_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_new_ones_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_new_zeros_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_adaptive_avg_pool3d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_avg_pool2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_avg_pool3d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_channel_shuffle_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_conv_transpose1d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_conv_transpose2d_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_cosine_embedding_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_dropout3d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_gelu_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_hardswish_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_huber_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_interpolate_bicubic_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_interpolate_linear_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_l1_loss_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_layer_norm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_linear_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_mse_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_pad_constant_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_pad_reflect_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_pad_replicate_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_pad_replicate_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_pad_replicate_negative_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_pixel_shuffle_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_rms_norm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_scaled_dot_product_attention_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_smooth_l1_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_softmin_with_dtype_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_softsign_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_unfold_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nn_functional_upsample_nearest_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nonzero_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_nonzero_static_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_norm_fro_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_norm_nuc_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_normal_in_place_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_outer_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_permute_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_polygamma_polygamma_n_1_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_polygamma_polygamma_n_4_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_pow_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_quantile_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_randn_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_randn_like_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_ravel_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_resize_as__cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_rsqrt_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_scatter_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_scatter_reduce_amax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_scatter_reduce_amin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_scatter_reduce_mean_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_scatter_reduce_sum_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_sgn_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_signal_windows_bartlett_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_signal_windows_general_cosine_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_signal_windows_hann_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_signal_windows_nuttall_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_sin_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_sinh_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_slice_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_slice_scatter_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_special_chebyshev_polynomial_u_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_special_erfcx_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_special_laguerre_polynomial_l_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_special_modified_bessel_k0_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_special_ndtri_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_special_scaled_modified_bessel_k1_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_special_xlog1py_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_split_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_split_list_args_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_square_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_squeeze_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_stack_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_std_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_std_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_std_unbiased_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_sum_to_size_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_t_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_t_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_t_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_take_along_dim_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_take_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_tan_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_tanh_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_tile_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_trace_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_transpose_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_transpose_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_tril_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_trunc_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_unfold_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_unsafe_split_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_unsqueeze_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_unsqueeze_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_var_mean_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_var_unbiased_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_view_as_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_view_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_vsplit_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_zero__cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_zeros_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_fail_gradgrad_zeros_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_NumpyCatCustomOp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_NumpyNonzeroCustomOp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_T_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad___radd___cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad___rmatmul___cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad___rmul___cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad___rpow___cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad___rpow___cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad__chunk_cat_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad__chunk_cat_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad__native_batch_norm_legit_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad__segment_reduce_offsets_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad__unsafe_masked_index_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_acosh_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_add_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_add_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_addbmm_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_addbmm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_addcdiv_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_addmm_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_addmm_decomposed_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_addr_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_alias_copy_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_allclose_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_angle_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_argsort_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_as_strided_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_as_strided_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_asin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_atan2_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_atan_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_atanh_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_atleast_2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_auto_functionalize_simple_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_bernoulli_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_block_diag_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_ceil_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_cfloat_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_cholesky_inverse_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_chunk_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_complex_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_cos_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_cosh_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_cov_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_cross_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_cumsum_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_cumsum_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_diag_embed_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_diagflat_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_diagonal_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_diagonal_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_diagonal_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_diff_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_dist_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_double_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_einsum_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_einsum_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_empty_like_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_empty_permuted_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_erfc_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_erfinv_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_expand_as_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_expand_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_expand_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_fft_fftshift_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_fft_fftshift_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_fft_hfft_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_fft_hfftn_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_fft_ifft_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_fft_ifftn_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_fft_ifftshift_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_fft_ifftshift_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_fft_ihfft2_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_fft_irfft_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_flatten_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_flipud_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_floor_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_floor_divide_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_fmax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_frexp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_full_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_gather_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_gather_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_gt_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_heaviside_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_hypot_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_igamma_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_imag_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_index_copy_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_index_reduce_mean_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_index_select_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_inner_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_int_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_isclose_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_isinf_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_isnan_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_isreal_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_istft_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_jiterator_4inputs_with_extra_args_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_jiterator_binary_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_jiterator_binary_return_by_ref_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_jiterator_binary_return_by_ref_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_kron_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_ldexp_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_le_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_lerp_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_lerp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_lgamma_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_cholesky_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_cond_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_det_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_det_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_det_singular_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_diagonal_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_eigh_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_eigvals_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_eigvalsh_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_inv_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_ldl_solve_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_ldl_solve_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_lstsq_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_lu_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_lu_factor_ex_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_lu_solve_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_matrix_power_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_matrix_rank_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_pinv_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_pinv_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_pinv_hermitian_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_qr_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_qr_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_solve_ex_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_solve_triangular_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_svd_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_svdvals_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_tensorsolve_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_tensorsolve_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_linalg_vander_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_logaddexp2_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_logdet_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_logical_or_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_mT_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_mT_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_map_triple_nested_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_masked_argmin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_masked_logaddexp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_masked_mean_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_masked_median_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_masked_norm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_masked_normalize_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_masked_prod_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_masked_select_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_masked_softmin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_masked_var_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_matmul_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_max_binary_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_max_reduction_no_dim_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_min_reduction_no_dim_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_mv_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nansum_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_narrow_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_native_layer_norm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_ne_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_new_empty_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_new_empty_strided_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_new_full_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_new_ones_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_adaptive_avg_pool1d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_adaptive_max_pool2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_binary_cross_entropy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_binary_cross_entropy_with_logits_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_conv1d_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_conv2d_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_conv3d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_conv_transpose2d_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_conv_transpose3d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_cosine_similarity_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_feature_alpha_dropout_without_train_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_hardsigmoid_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_interpolate_bicubic_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_interpolate_linear_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_interpolate_nearest_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_layer_norm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_linear_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_local_response_norm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_max_pool2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_max_pool3d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_max_unpool1d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_max_unpool1d_grad_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_max_unpool2d_grad_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_max_unpool3d_grad_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_multi_margin_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_nll_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_pad_constant_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_pad_replicate_negative_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_pairwise_distance_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_silu_complex_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_soft_margin_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_tanhshrink_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_tanhshrink_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_triplet_margin_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_triplet_margin_with_distance_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_unfold_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nn_functional_upsample_nearest_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_nonzero_static_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_norm_nuc_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_normal_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_normal_in_place_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_normal_number_mean_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_ones_like_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_ormqr_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_outer_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_pca_lowrank_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_pca_lowrank_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_permute_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_pinverse_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_pinverse_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_polygamma_polygamma_n_0_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_positive_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_pow_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_quantile_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_rand_like_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_randint_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_randint_like_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_randn_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_ravel_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_repeat_interleave_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_reshape_as_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_resize__cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_resize_as__cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_scatter_add_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_scatter_reduce_amin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_scatter_reduce_mean_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_searchsorted_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_select_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_sgn_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_sigmoid_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_signal_windows_general_cosine_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_slice_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_softmax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_bessel_j0_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_bessel_y1_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_chebyshev_polynomial_u_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_erfcx_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_i0e_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_log_ndtr_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_modified_bessel_i1_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_modified_bessel_k0_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_ndtri_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_polygamma_special_polygamma_n_0_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_scaled_modified_bessel_k0_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_shifted_chebyshev_polynomial_t_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_shifted_chebyshev_polynomial_u_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_xlog1py_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_special_zeta_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_split_list_args_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_split_with_sizes_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_square_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_square_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_squeeze_multiple_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_squeeze_multiple_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_stack_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_std_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_std_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_std_mean_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_std_mean_unbiased_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_stft_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_sum_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_svd_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_svd_lowrank_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_take_along_dim_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_tensor_split_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_tensordot_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_to_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_true_divide_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_unflatten_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_unfold_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_unique_consecutive_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_unsafe_chunk_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_unsqueeze_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_vdot_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_vdot_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_view_as_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_view_as_real_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_view_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_view_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_vsplit_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_grad_where_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_H_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_NumpyCubeCustomOp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_NumpyNonzeroCustomOp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_NumpySplitCopyWithIntCustomOp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_NumpyTakeCustomOp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad___getitem___cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad___rmod___cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad__unsafe_masked_index_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad__unsafe_masked_index_put_accumulate_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_acosh_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_acosh_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_addcdiv_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_addmm_decomposed_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_addr_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_addr_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_alias_copy_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_amax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_amin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_any_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_argmax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_argsort_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_argwhere_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_as_strided_copy_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_as_strided_partial_views_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_atleast_2d_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_atleast_3d_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_block_diag_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_broadcast_tensors_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_cdist_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_cdouble_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_chalf_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_cholesky_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_cholesky_solve_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_cholesky_solve_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_chunk_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_clone_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_cond_simple_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_conj_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_conj_physical_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_contiguous_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_cos_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_count_nonzero_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_cov_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_cummax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_cummin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_cumprod_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_cumsum_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_diag_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_diag_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_diag_embed_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_diagonal_copy_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_diagonal_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_diagonal_scatter_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_div_floor_rounding_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_div_trunc_rounding_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_dot_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_dsplit_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_empty_like_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_equal_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_erfc_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_exp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_expand_as_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_eye_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_fft_fft2_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_fft_fftshift_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_fft_hfft2_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_fft_hfft2_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_fft_ifft_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_fft_irfft2_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_fill_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_flatten_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_float_power_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_floor_divide_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_gather_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_geometric_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_geqrf_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_gradient_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_gradient_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_grid_sampler_2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_hsplit_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_hstack_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_hypot_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_i0_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_index_add_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_index_copy_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_index_fill_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_index_put_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_index_reduce_amax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_index_reduce_amin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_index_reduce_mean_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_index_reduce_prod_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_index_select_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_isclose_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_isinf_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_isreal_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_jiterator_2inputs_2outputs_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_jiterator_binary_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_kron_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_kron_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_cond_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_cond_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_cross_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_cross_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_det_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_det_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_det_singular_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_eig_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_eigvalsh_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_inv_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_inv_ex_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_inv_ex_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_ldl_factor_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_ldl_factor_ex_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_lu_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_lu_solve_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_matrix_power_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_matrix_rank_hermitian_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_multi_dot_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_norm_subgradients_at_zero_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_pinv_singular_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_slogdet_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_svd_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_svdvals_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_svdvals_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_linalg_tensorsolve_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_log10_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_log1p_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_log2_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_log_softmax_with_dtype_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_logaddexp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_logcumsumexp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_logdet_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_logical_and_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_logical_or_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_logspace_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_logspace_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_long_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_mT_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_map_simple_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_masked_amax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_masked_argmin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_masked_cumprod_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_masked_fill_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_masked_normalize_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_masked_normalize_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_masked_softmax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_masked_sum_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_masked_var_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_matmul_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_matrix_exp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_max_reduction_no_dim_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_min_binary_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_mv_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nan_to_num_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_narrow_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_ne_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_ne_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_new_empty_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_new_ones_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_new_ones_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_adaptive_avg_pool1d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_alpha_dropout_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_avg_pool1d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_conv2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_conv3d_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_conv_transpose3d_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_conv_transpose3d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_dropout2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_feature_alpha_dropout_without_train_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_feature_alpha_dropout_without_train_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_gelu_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_glu_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_hinge_embedding_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_interpolate_bicubic_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_kl_div_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_l1_loss_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_leaky_relu_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_linear_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_max_pool2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_max_unpool1d_grad_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_max_unpool2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_max_unpool2d_grad_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_multi_head_attention_forward_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_multilabel_soft_margin_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_pad_circular_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_pad_replicate_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_pairwise_distance_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_pixel_unshuffle_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_prelu_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_relu6_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_smooth_l1_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_softmin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_softmin_with_dtype_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_triplet_margin_with_distance_loss_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_triplet_margin_with_distance_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_unfold_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_nn_functional_upsample_bilinear_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_norm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_norm_nuc_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_norm_nuc_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_normal_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_normal_number_mean_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_outer_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_outer_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_pinverse_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_polar_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_polygamma_polygamma_n_3_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_pow_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_quantile_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_randint_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_real_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_real_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_reciprocal_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_repeat_interleave_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_reshape_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_resize_as__cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_resolve_neg_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_round_decimals_3_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_round_decimals_neg_3_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_scatter_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_scatter_reduce_prod_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_scatter_reduce_sum_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_sgn_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_sigmoid_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_signal_windows_gaussian_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_signal_windows_general_hamming_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_sin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_sinc_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_slice_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_softmax_with_dtype_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_softmax_with_dtype_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_special_bessel_j0_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_special_chebyshev_polynomial_u_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_special_chebyshev_polynomial_w_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_special_entr_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_special_i0e_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_special_scaled_modified_bessel_k1_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_special_shifted_chebyshev_polynomial_u_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_special_shifted_chebyshev_polynomial_v_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_special_zeta_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_split_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_squeeze_multiple_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_stack_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_std_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_std_unbiased_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_std_unbiased_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_sub_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_svd_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_take_along_dim_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_tan_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_tanh_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_tensordot_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_to_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_to_sparse_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_torch_ops_aten__safe_softmax_default_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_trace_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_trace_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_transpose_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_trapz_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_triu_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_true_divide_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_trunc_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_uniform_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_unsafe_chunk_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_unsafe_split_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_unsqueeze_copy_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_var_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_vdot_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_view_as_complex_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_view_as_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_vsplit_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_vstack_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_where_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_zeros_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_fn_gradgrad_zeros_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_NumpyNMSCustomOp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_NumpySplitCopyWithIntCustomOp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_NumpyTakeCustomOp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad___rmul___cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad___rsub___cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad__native_batch_norm_legit_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad__segment_reduce_lengths_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad__unsafe_masked_index_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad__unsafe_masked_index_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad__unsafe_masked_index_put_accumulate_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_addcdiv_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_addcmul_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_addmm_decomposed_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_addmm_decomposed_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_addmv_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_alias_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_allclose_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_angle_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_argmax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_argmin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_as_strided_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_as_strided_scatter_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_as_strided_scatter_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_atan_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_atleast_1d_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_atleast_2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_baddbmm_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_bmm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_broadcast_to_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_broadcast_to_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_cat_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_cauchy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_cdouble_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_cdouble_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_char_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_cholesky_inverse_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_chunk_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_clamp_min_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_clone_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_clone_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_column_stack_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_combinations_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_constant_pad_nd_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_constant_pad_nd_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_cov_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_cummax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_cumprod_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_cumsum_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_cumsum_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_cumulative_trapezoid_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_deg2rad_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_diag_embed_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_diagflat_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_diagonal_scatter_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_digamma_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_div_no_rounding_mode_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_dsplit_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_dstack_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_empty_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_empty_like_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_empty_strided_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_eq_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_equal_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_expand_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_expand_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_exponential_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_eye_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_fft_fft2_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_fft_fft_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_fft_fftn_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_fft_ihfft2_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_fft_irfft_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_fft_irfftn_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_fft_rfft2_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_fill_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_fill_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_flatten_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_fliplr_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_float_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_floor_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_fmin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_full_like_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_geqrf_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_gradient_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_half_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_heaviside_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_histc_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_index_add_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_index_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_index_put_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_index_select_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_inner_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_inner_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_isin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_isnan_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_isnan_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_jiterator_2inputs_2outputs_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_jiterator_4inputs_with_extra_args_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_jiterator_unary_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_kthvalue_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_lerp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_cholesky_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_cross_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_det_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_eig_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_householder_product_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_lstsq_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_lu_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_lu_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_lu_factor_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_lu_factor_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_lu_factor_ex_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_matrix_rank_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_matrix_rank_hermitian_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_norm_subgradients_at_zero_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_norm_subgradients_at_zero_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_pinv_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_pinv_hermitian_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_pinv_singular_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_pinv_singular_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_solve_ex_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_solve_ex_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_solve_triangular_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_svdvals_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_tensorinv_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linalg_vecdot_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linspace_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_linspace_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_log10_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_log1p_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_log1p_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_log_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_log_softmax_with_dtype_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_logaddexp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_logical_or_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_logical_or_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_logspace_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_logspace_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_long_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_lt_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_lu_solve_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_masked_argmax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_masked_scatter_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_masked_std_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_matmul_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_max_reduction_with_dim_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_meshgrid_variadic_tensors_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_min_reduction_no_dim_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_mm_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_movedim_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_multinomial_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nansum_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nansum_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_narrow_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_native_dropout_backward_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_new_empty_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_new_empty_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_new_empty_strided_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_adaptive_max_pool1d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_adaptive_max_pool3d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_alpha_dropout_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_avg_pool2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_avg_pool3d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_batch_norm_without_cudnn_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_binary_cross_entropy_with_logits_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_conv1d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_conv_transpose2d_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_cosine_similarity_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_dropout3d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_fractional_max_pool2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_gaussian_nll_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_glu_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_hardshrink_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_hinge_embedding_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_interpolate_trilinear_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_linear_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_margin_ranking_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_max_pool1d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_max_unpool2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_max_unpool3d_grad_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_mish_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_multilabel_margin_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_nll_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_normalize_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_pad_circular_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_pad_constant_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_pad_reflect_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_pairwise_distance_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_pixel_shuffle_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_pixel_unshuffle_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_poisson_nll_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_rms_norm_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_softmin_with_dtype_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_tanhshrink_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nn_functional_triplet_margin_with_distance_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_nonzero_static_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_norm_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_norm_fro_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_norm_fro_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_norm_inf_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_norm_nuc_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_norm_nuc_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_normal_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_normal_in_place_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_normal_number_mean_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_ones_like_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_put_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_randint_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_randn_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_real_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_real_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_repeat_interleave_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_reshape_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_resize__cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_round_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_round_decimals_0_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_round_decimals_3_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_round_decimals_neg_3_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_rsqrt_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_scatter_reduce_amax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_scatter_reduce_mean_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_scatter_reduce_sum_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_select_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_select_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_select_scatter_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_sgn_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_short_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_sigmoid_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_signal_windows_exponential_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_signal_windows_general_cosine_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_signal_windows_kaiser_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_signal_windows_nuttall_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_sinh_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_softmax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_special_bessel_y1_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_special_erfcx_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_special_hermite_polynomial_h_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_special_legendre_polynomial_p_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_split_with_sizes_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_square_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_squeeze_multiple_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_squeeze_multiple_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_stack_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_std_unbiased_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_sub_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_sub_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_sum_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_sum_to_size_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_take_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_tensordot_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_tile_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_to_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_trace_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_trapezoid_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_trapz_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_triangular_solve_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_unbind_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_unflatten_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_uniform_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_uniform_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_unsafe_chunk_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_unsqueeze_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_var_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_view_copy_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_view_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_vsplit_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_grad_zeros_like_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_T_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad___getitem___cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad___radd___cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad___rmul___cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad__chunk_cat_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad__unsafe_masked_index_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_addcdiv_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_addcdiv_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_addcmul_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_addmm_decomposed_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_alias_copy_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_allclose_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_arange_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_argmin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_argwhere_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_as_strided_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_as_strided_scatter_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_asin_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_asinh_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_atan2_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_atan_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_bernoulli_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_bfloat16_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_broadcast_tensors_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_broadcast_tensors_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_cat_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_cat_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_cdist_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_cdouble_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_chalf_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_char_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_cholesky_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_cholesky_inverse_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_cholesky_solve_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_chunk_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_column_stack_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_combinations_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_complex_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_conj_physical_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_constant_pad_nd_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_corrcoef_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_cos_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_count_nonzero_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_cov_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_cross_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_cummax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_cumsum_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_diagonal_scatter_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_diff_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_div_no_rounding_mode_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_div_no_rounding_mode_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_div_trunc_rounding_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_dsplit_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_empty_like_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_empty_permuted_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_erfinv_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_exp_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_exp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_expand_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_expand_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_eye_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_fft_fft2_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_fft_fft_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_fft_fftn_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_fft_fftn_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_fft_fftshift_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_fft_hfft2_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_fft_hfft_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_fft_ifft2_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_fft_ifftn_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_fft_ihfft_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_fft_ihfftn_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_fft_irfft2_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_flip_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_fliplr_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_float_power_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_frexp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_geometric_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_gradient_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_gt_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_histc_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_igammac_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_imag_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_index_reduce_amin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_int_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_isfinite_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_isnan_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_isnan_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_jiterator_2inputs_2outputs_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_jiterator_4inputs_with_extra_args_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_jiterator_binary_return_by_ref_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_jiterator_binary_return_by_ref_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_jiterator_unary_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_lerp_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_cholesky_ex_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_cond_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_cross_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_diagonal_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_eigvalsh_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_eigvalsh_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_householder_product_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_ldl_factor_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_ldl_factor_ex_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_lstsq_grad_oriented_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_lu_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_lu_factor_ex_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_lu_solve_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_matrix_rank_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_pinv_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_qr_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_qr_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_slogdet_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_solve_ex_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_svdvals_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_vecdot_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_vector_norm_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_linalg_vector_norm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_log10_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_log_softmax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_logaddexp2_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_logical_and_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_logical_and_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_logit_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_logsumexp_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_long_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_lu_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_lu_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_lu_solve_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_mT_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_masked_argmin_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_masked_fill_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_masked_prod_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_masked_select_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_masked_softmax_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_masked_std_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_matrix_exp_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_max_binary_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_max_reduction_with_dim_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_meshgrid_list_of_tensors_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_meshgrid_list_of_tensors_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_meshgrid_variadic_tensors_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_min_reduction_with_dim_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_movedim_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_mvlgamma_mvlgamma_p_1_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_narrow_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_narrow_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_native_batch_norm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_new_empty_strided_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_new_full_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_new_zeros_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_adaptive_avg_pool3d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_avg_pool3d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_binary_cross_entropy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_celu_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_conv1d_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_conv2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_cosine_similarity_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_dropout2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_dropout_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_embedding_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_fractional_max_pool2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_group_norm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_interpolate_bicubic_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_interpolate_bilinear_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_interpolate_nearest-exact_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_layer_norm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_leaky_relu_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_max_pool2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_max_pool3d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_max_unpool1d_grad_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_max_unpool2d_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_max_unpool2d_grad_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_max_unpool3d_grad_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_mish_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_multi_head_attention_forward_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_normalize_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_pad_constant_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_pad_constant_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_pad_replicate_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_pad_replicate_negative_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_pad_replicate_negative_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_pairwise_distance_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_pdist_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_pixel_unshuffle_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_relu_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_silu_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_softmin_with_dtype_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_softmin_with_dtype_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_softplus_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_softshrink_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_softsign_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_triplet_margin_loss_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_triplet_margin_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nn_functional_triplet_margin_with_distance_loss_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nonzero_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_nonzero_static_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_norm_nuc_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_ones_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_outer_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_permute_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_pinverse_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_polar_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_polygamma_polygamma_n_2_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_positive_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_positive_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_put_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_qr_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_rand_like_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_real_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_renorm_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_repeat_interleave_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_reshape_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_roll_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_round_decimals_neg_3_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_scalar_tensor_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_scatter_add_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_scatter_reduce_mean_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_select_scatter_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_signal_windows_exponential_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_signal_windows_hamming_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_special_bessel_y0_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_special_chebyshev_polynomial_w_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_special_entr_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_special_hermite_polynomial_he_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_special_i1e_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_special_modified_bessel_i1_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_special_modified_bessel_k1_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_special_shifted_chebyshev_polynomial_t_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_special_shifted_chebyshev_polynomial_w_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_special_zeta_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_split_with_sizes_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_square_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_stack_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_stack_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_std_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_std_mean_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_std_mean_unbiased_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_sum_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_take_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_take_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_tan_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_tan_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_tensor_split_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_tensordot_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_to_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_torch_ops_aten__safe_softmax_default_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_transpose_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_trapezoid_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_triangular_solve_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_tril_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_triu_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_trunc_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_unbind_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_unfold_copy_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_unfold_copy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_var_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_var_mean_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_var_mean_unbiased_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_var_mean_unbiased_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_view_as_real_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_view_copy_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_vsplit_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_where_cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_xlogy_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_zero__cuda_complex128, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_zeros_cuda_float64, test/test_ops_gradients.py::TestBwdGradientsCUDA::test_inplace_gradgrad_zeros_like_cuda_complex128 2024-08-06T21:33:38.3594303Z 2024-08-06T21:33:41.8840690Z Running test_ops 2/11 ... [2024-08-06 21:33:41.883532] 2024-08-06T21:33:41.8843995Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_ops.py', '-m', 'not serial', '--shard-id=2', '--num-shards=11', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:33:41.883986] 2024-08-06T21:38:01.8640104Z 2024-08-06T21:38:01.8641028Z test_meta 5/5 was successful, full logs can be found in artifacts with path test/test-reports/test_meta_5.5_2903666c02d5e4b6_.log 2024-08-06T21:38:02.1935098Z Running 8161 items in this shard: test/test_meta.py::TestMetaConverter::test_channels_last_non_leaf, test/test_meta.py::TestMetaConverter::test_tensor_outlives_converter, test/test_meta.py::TestMetaConverter::test_view_of_leaf, test/test_meta.py::TestMetaCUDA::test_batch_norm_backward_output_mask3_cuda, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype___rmul___cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype___rsub___cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype__refs_atan2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype__refs_div_floor_rounding_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype__refs_eq_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype__refs_igamma_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype__refs_minimum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype__refs_ne_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype__refs_remainder_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype__refs_xlogy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype_atan2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype_div_floor_rounding_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype_fmin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype_fmod_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype_logaddexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype_logical_or_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype_min_binary_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype_ne_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype_nextafter_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype_special_chebyshev_polynomial_v_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype_special_zeta_cuda_float32, test/test_meta.py::TestMetaCUDA::test_binary_ufuncs_mixed_dtype_sub_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_H_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_T_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___getitem___cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___radd___cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___radd___cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___radd___cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rand___cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rand___cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rdiv___cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rdiv___cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rdiv___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rmul___cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rmul___cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rpow___cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rsub___cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rsub___cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rsub___cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rsub___cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rsub___cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rxor___cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace___rxor___cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__chunk_cat_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__chunk_cat_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__chunk_cat_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__chunk_cat_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__chunk_cat_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__chunk_cat_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_abs_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_abs_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_abs_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_acos_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_acos_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_addcdiv_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_addcdiv_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_addcdiv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_addcdiv_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_asin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_asin_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_asin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_asin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_atan_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_ceil_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_clamp_max_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_clamp_max_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_clamp_max_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_clamp_max_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_clamp_min_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_copy_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_cos_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_cosh_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_cosh_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_cosh_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_cosh_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_cosh_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_div_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_div_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_div_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_erf_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_erfc_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_erfc_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_erfc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_erfc_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_exp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_exp_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_expm1_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_expm1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_floor_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_floor_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_floor_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_floor_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_floor_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_frac_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_frac_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_lerp_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_lerp_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_lgamma_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_lgamma_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_log10_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_log10_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_log1p_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_log1p_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_log1p_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_log1p_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_log2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_log2_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_log2_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_log_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_log_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_max_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_maximum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_maximum_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_minimum_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_mul_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_mul_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_mul_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_neg_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_neg_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_neg_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_pow_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_reciprocal_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_round_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_round_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_round_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sigmoid_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sign_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sign_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sign_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sinh_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sqrt_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sqrt_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sqrt_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sqrt_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sqrt_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sqrt_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sub_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sub_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_sub_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_tan_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_tan_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_tan_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_tan_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_tanh_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_tanh_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_trunc_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_trunc_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_trunc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_trunc_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_trunc_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_trunc_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_trunc_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_zero_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__foreach_zero_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__native_batch_norm_legit_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__native_batch_norm_legit_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__softmax_backward_data_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__unsafe_masked_index_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__unsafe_masked_index_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__unsafe_masked_index_put_accumulate_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__unsafe_masked_index_put_accumulate_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__unsafe_masked_index_put_accumulate_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace__upsample_bilinear2d_aa_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_abs_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_abs_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_acos_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_acos_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_acosh_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_acosh_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_acosh_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_acosh_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_add_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_add_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_addbmm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_addcmul_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_addmm_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_addmm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_addmm_decomposed_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_addmv_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_addmv_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_addmv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_addr_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_alias_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_alias_copy_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_alias_copy_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_all_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_all_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_all_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_allclose_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_amax_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_amax_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_amin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_amin_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_aminmax_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_angle_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_angle_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_any_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_any_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_any_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_any_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_argmax_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_argmax_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_argsort_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_argsort_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_argsort_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_argsort_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_argsort_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_argwhere_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_argwhere_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_as_strided_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_as_strided_copy_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_as_strided_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_as_strided_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_as_strided_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_as_strided_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_as_strided_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_as_strided_partial_views_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_as_strided_scatter_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_as_strided_scatter_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_as_strided_scatter_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_asin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_asin_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_asin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_asin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_asin_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_asinh_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_asinh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_asinh_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atan2_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atanh_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atanh_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atanh_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atleast_1d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atleast_1d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atleast_2d_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atleast_2d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atleast_2d_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atleast_2d_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atleast_2d_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atleast_3d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atleast_3d_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atleast_3d_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_atleast_3d_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_baddbmm_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bernoulli_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bfloat16_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bfloat16_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bfloat16_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bfloat16_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bitwise_and_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bitwise_and_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bitwise_left_shift_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bitwise_left_shift_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bitwise_not_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bitwise_not_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bitwise_or_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bitwise_right_shift_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bitwise_xor_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bitwise_xor_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_block_diag_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_block_diag_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_block_diag_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bmm_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bool_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bool_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_broadcast_tensors_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_broadcast_to_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_broadcast_to_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_bucketize_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_byte_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_byte_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_byte_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_byte_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cartesian_prod_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cartesian_prod_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cat_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cdist_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cdouble_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cdouble_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ceil_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ceil_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ceil_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cfloat_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_chalf_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_chalf_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_char_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_char_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cholesky_inverse_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cholesky_solve_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cholesky_solve_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_chunk_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_chunk_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_clamp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_clamp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_clamp_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_clamp_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_clamp_min_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_clamp_min_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_clamp_min_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_clamp_min_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_clamp_min_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_clone_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_clone_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_column_stack_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_column_stack_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_column_stack_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_combinations_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_conj_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_conj_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_conj_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_conj_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_conj_physical_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_conj_physical_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_constant_pad_nd_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_constant_pad_nd_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_constant_pad_nd_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_contiguous_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_copysign_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_copysign_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cosh_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cosh_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cov_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cov_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cov_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cov_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cummax_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cummax_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cummax_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cumprod_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cumprod_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cumprod_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cumsum_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cumsum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cumulative_trapezoid_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_cumulative_trapezoid_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_deg2rad_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_deg2rad_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_diag_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_diag_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_diag_embed_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_diagflat_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_diagonal_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_diagonal_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_diagonal_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_diagonal_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_diagonal_scatter_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_diagonal_scatter_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_diagonal_scatter_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_diagonal_scatter_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_diff_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_diff_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_digamma_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_dist_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_div_floor_rounding_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_div_floor_rounding_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_div_no_rounding_mode_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_div_no_rounding_mode_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_div_no_rounding_mode_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_div_trunc_rounding_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_div_trunc_rounding_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_div_trunc_rounding_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_dot_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_double_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_double_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_dsplit_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_dsplit_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_dsplit_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_dsplit_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_dstack_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_dstack_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_dstack_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_einsum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_empty_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_empty_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_empty_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_empty_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_empty_like_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_empty_like_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_empty_permuted_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_empty_permuted_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_empty_permuted_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_empty_strided_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_empty_strided_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_empty_strided_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_eq_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_eq_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_equal_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_equal_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_erf_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_erfc_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_erfc_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_erfc_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_exp2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_exp2_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_exp_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_exp_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_exp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_exp_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_exp_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_exp_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_exp_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_expand_as_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_expand_as_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_expand_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_expand_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_expand_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_expand_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_expand_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_expand_copy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_expand_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_expm1_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_expm1_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_expm1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_exponential_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_eye_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_eye_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_fft_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_fft_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_fftn_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_fftn_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_fftn_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_fftshift_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_fftshift_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_fftshift_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_hfft2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_hfft2_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_hfft2_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_hfft_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_hfft_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_hfftn_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_hfftn_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_ifft2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_ifft2_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_ifft2_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_ifft_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_ifftn_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_ifftshift_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_ifftshift_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_ifftshift_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_ihfft2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_ihfft2_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_ihfft_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_ihfft_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_ihfftn_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_irfft2_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_irfft2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_irfft2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_irfft2_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_irfft_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_irfftn_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_rfft2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_rfft_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_rfftn_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_rfftn_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fft_rfftn_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fill_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_flatten_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_flatten_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_flip_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fliplr_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fliplr_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fliplr_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_flipud_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_flipud_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_flipud_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_flipud_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_flipud_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_float_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_float_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_float_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_float_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_float_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_float_power_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_float_power_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_float_power_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_float_power_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_floor_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_floor_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_floor_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_floor_divide_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fmin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fmod_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fmod_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_fmod_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_full_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_full_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_full_like_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_full_like_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_full_like_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_gather_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_gather_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_gcd_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_gcd_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_gcd_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_gcd_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ge_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ge_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_geometric_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_geometric_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_gradient_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_gradient_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_gradient_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_gradient_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_grid_sampler_2d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_gt_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_gt_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_half_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_half_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_half_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_histc_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_hsplit_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_hsplit_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_hsplit_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_hstack_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_hstack_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_i0_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_imag_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_add_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_add_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_add_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_copy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_fill_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_fill_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_fill_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_fill_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_fill_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_put_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_put_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_reduce_amax_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_reduce_amax_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_reduce_amin_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_reduce_amin_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_reduce_mean_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_reduce_prod_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_reduce_prod_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_reduce_prod_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_select_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_index_select_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_inner_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_inner_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_int_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_int_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isclose_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isfinite_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isfinite_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isinf_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isinf_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isinf_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isinf_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isnan_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isnan_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isnan_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isnan_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isposinf_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isposinf_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_isreal_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_item_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_jiterator_2inputs_2outputs_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_jiterator_2inputs_2outputs_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_jiterator_4inputs_with_extra_args_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_jiterator_4inputs_with_extra_args_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_jiterator_binary_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_jiterator_binary_return_by_ref_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_jiterator_binary_return_by_ref_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_jiterator_binary_return_by_ref_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_jiterator_unary_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_jiterator_unary_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_kron_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_kthvalue_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_lcm_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ldexp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ldexp_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ldexp_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ldexp_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ldexp_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_le_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_lerp_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_lerp_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_lgamma_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_lgamma_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_cholesky_ex_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_cholesky_ex_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_cholesky_ex_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_cross_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_cross_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_det_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_diagonal_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_diagonal_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_diagonal_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_eig_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_eig_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_eigh_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_eigh_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_eigh_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_eigvals_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_eigvalsh_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_inv_ex_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_ldl_factor_ex_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_ldl_solve_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_lstsq_grad_oriented_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_lstsq_grad_oriented_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_lu_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_lu_factor_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_lu_factor_ex_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_matrix_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_matrix_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_matrix_power_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_matrix_rank_hermitian_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_multi_dot_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_pinv_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_pinv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_pinv_singular_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_slogdet_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_solve_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_solve_triangular_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_svdvals_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_tensorinv_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_tensorinv_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_tensorsolve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_vander_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_vander_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_vander_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_vecdot_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_vecdot_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linalg_vecdot_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linspace_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linspace_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linspace_tensor_overload_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_linspace_tensor_overload_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log10_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log10_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log10_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log10_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log10_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log1p_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log2_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log2_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log2_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log_normal_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log_softmax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log_softmax_with_dtype_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log_softmax_with_dtype_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log_softmax_with_dtype_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log_softmax_with_dtype_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_log_softmax_with_dtype_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logaddexp2_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logaddexp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logaddexp_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logaddexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logaddexp_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logdet_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logical_and_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logical_and_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logical_not_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logical_or_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logical_or_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logical_or_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logical_xor_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logical_xor_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logical_xor_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logit_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logit_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logit_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logit_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logit_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logspace_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logspace_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logspace_tensor_overload_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logspace_tensor_overload_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logsumexp_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logsumexp_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_logsumexp_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_long_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_long_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_long_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_long_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_lt_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_lt_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_lu_solve_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_lu_unpack_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mH_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mT_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mT_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mT_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mT_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mT_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_amin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_amin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_argmax_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_cumprod_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_cumprod_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_cumprod_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_cumprod_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_cumprod_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_cumsum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_cumsum_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_cumsum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_fill_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_log_softmax_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_logaddexp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_logaddexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_logsumexp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_logsumexp_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_mean_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_mean_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_normalize_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_prod_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_prod_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_prod_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_scatter_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_scatter_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_select_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_softmax_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_softmin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_std_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_std_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_std_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_sum_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_sum_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_masked_var_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_matrix_exp_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_matrix_exp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_max_binary_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_max_binary_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_max_binary_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_max_reduction_no_dim_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_max_reduction_no_dim_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_max_reduction_with_dim_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_maximum_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_maximum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mean_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mean_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mean_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_median_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_median_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_median_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_meshgrid_variadic_tensors_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_meshgrid_variadic_tensors_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_min_binary_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_min_reduction_no_dim_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_min_reduction_with_dim_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_minimum_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mode_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mode_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mode_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_movedim_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_movedim_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_movedim_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_movedim_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_msort_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_msort_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mul_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mul_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mul_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_mvlgamma_mvlgamma_p_5_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nan_to_num_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nan_to_num_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nanmean_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nansum_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nansum_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nansum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nansum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nansum_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_narrow_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_narrow_copy_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_narrow_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_narrow_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_narrow_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_native_batch_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ne_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ne_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ne_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_neg_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_neg_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_neg_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_neg_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_new_empty_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_new_empty_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_new_empty_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_new_empty_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_new_full_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_new_ones_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_new_ones_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_new_ones_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_new_zeros_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_new_zeros_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_new_zeros_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_new_zeros_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nextafter_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_adaptive_avg_pool2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_adaptive_avg_pool2d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_adaptive_max_pool2d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_avg_pool2d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_batch_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_batch_norm_without_cudnn_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_batch_norm_without_cudnn_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_binary_cross_entropy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_binary_cross_entropy_with_logits_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_binary_cross_entropy_with_logits_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_channel_shuffle_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_channel_shuffle_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_conv1d_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_conv1d_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_conv1d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_conv2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_conv2d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_conv3d_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_conv3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_conv_transpose1d_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_conv_transpose1d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_conv_transpose3d_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_cosine_embedding_loss_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_cosine_embedding_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_cosine_embedding_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_cosine_embedding_loss_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_cross_entropy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_dropout2d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_dropout2d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_dropout3d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_dropout_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_dropout_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_dropout_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_embedding_bag_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_embedding_bag_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_feature_alpha_dropout_without_train_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_feature_alpha_dropout_without_train_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_feature_alpha_dropout_without_train_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_feature_alpha_dropout_without_train_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_fractional_max_pool2d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_fractional_max_pool3d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_gaussian_nll_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_glu_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_group_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_hardshrink_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_hardshrink_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_hardshrink_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_hardtanh_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_hardtanh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_hardtanh_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_hinge_embedding_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_interpolate_bilinear_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_interpolate_nearest-exact_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_interpolate_nearest_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_interpolate_trilinear_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_kl_div_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_l1_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_layer_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_leaky_relu_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_linear_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_linear_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_linear_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_logsigmoid_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_margin_ranking_loss_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_margin_ranking_loss_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_max_pool1d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_max_pool3d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_max_unpool1d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_max_unpool1d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_max_unpool1d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_mish_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_mse_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_multi_head_attention_forward_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_multi_margin_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_multilabel_margin_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_multilabel_margin_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_nll_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_normalize_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pad_circular_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pad_circular_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pad_circular_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pad_constant_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pad_reflect_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pad_reflect_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pad_reflect_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pad_replicate_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pad_replicate_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pad_replicate_negative_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pairwise_distance_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pairwise_distance_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pairwise_distance_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pairwise_distance_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pixel_shuffle_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pixel_shuffle_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pixel_unshuffle_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pixel_unshuffle_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pixel_unshuffle_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_pixel_unshuffle_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_poisson_nll_loss_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_poisson_nll_loss_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_prelu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_prelu_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_relu6_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_relu6_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_rms_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_rms_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_rrelu_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_scaled_dot_product_attention_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_silu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_soft_margin_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_softmin_with_dtype_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_softmin_with_dtype_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_softplus_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_softshrink_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_softsign_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_softsign_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_softsign_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_tanhshrink_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_tanhshrink_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_triplet_margin_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_triplet_margin_loss_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_triplet_margin_loss_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_triplet_margin_with_distance_loss_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_triplet_margin_with_distance_loss_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_unfold_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_upsample_bilinear_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_upsample_nearest_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nn_functional_upsample_nearest_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nonzero_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nonzero_static_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nonzero_static_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nonzero_static_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_nonzero_static_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_norm_fro_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_norm_fro_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_norm_fro_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_norm_inf_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_norm_inf_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_norm_nuc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_norm_nuc_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_normal_in_place_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_normal_in_place_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_normal_number_mean_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_normal_number_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_normal_number_mean_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ones_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ones_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ones_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ones_like_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ormqr_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ormqr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_outer_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_outer_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_outer_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_pca_lowrank_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_pca_lowrank_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_permute_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_permute_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_permute_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_pinverse_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_pinverse_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_polar_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_polygamma_polygamma_n_0_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_polygamma_polygamma_n_0_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_polygamma_polygamma_n_1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_polygamma_polygamma_n_1_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_polygamma_polygamma_n_2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_polygamma_polygamma_n_3_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_polygamma_polygamma_n_3_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_polygamma_polygamma_n_4_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_polygamma_polygamma_n_4_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_positive_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_pow_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_pow_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_pow_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_pow_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_pow_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_prod_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_prod_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_prod_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_prod_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_put_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_put_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_put_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_qr_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_qr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_quantile_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_rand_like_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_randint_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_randint_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_randint_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_randint_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_randn_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_randn_like_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ravel_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_ravel_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_real_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_reciprocal_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_reciprocal_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_renorm_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_renorm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_repeat_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_repeat_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_repeat_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_repeat_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_repeat_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_repeat_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_repeat_interleave_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_repeat_interleave_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_repeat_interleave_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_reshape_as_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_reshape_as_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_reshape_as_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_reshape_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_reshape_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_resize__cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_resize__cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_resize__cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_resize_as__cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_resize_as__cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_resolve_conj_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_resolve_conj_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_roll_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_roll_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_rot90_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_rot90_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_rot90_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_rot90_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_round_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_round_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_round_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_round_decimals_0_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_round_decimals_3_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_round_decimals_3_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_rsqrt_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_rsqrt_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_rsqrt_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_rsqrt_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_rsqrt_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_rsub_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_rsub_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_scalar_tensor_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_scalar_tensor_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_scalar_tensor_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_scatter_add_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_scatter_add_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_scatter_add_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_scatter_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_scatter_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_scatter_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_scatter_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_scatter_reduce_prod_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_scatter_reduce_prod_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_scatter_reduce_sum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_searchsorted_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_searchsorted_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_select_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_select_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_select_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_select_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_select_scatter_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_select_scatter_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_select_scatter_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_select_scatter_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_select_scatter_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sgn_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_short_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_short_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_short_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_short_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sigmoid_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sign_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sign_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sign_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_signal_windows_blackman_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_signal_windows_gaussian_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_signal_windows_general_cosine_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_signbit_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_signbit_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_signbit_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sin_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sin_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sin_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sinc_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sinc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sinh_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sinh_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sinh_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sinh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sinh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sinh_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_slice_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_slice_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_slice_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_slice_scatter_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_slice_scatter_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_softmax_with_dtype_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sort_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sort_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sparse_mm_reduce_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sparse_sampled_addmm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_bessel_j0_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_bessel_j0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_bessel_y0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_bessel_y0_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_bessel_y0_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_bessel_y1_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_chebyshev_polynomial_t_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_chebyshev_polynomial_u_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_chebyshev_polynomial_v_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_chebyshev_polynomial_v_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_chebyshev_polynomial_v_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_chebyshev_polynomial_v_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_chebyshev_polynomial_w_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_chebyshev_polynomial_w_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_entr_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_entr_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_erfcx_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_hermite_polynomial_h_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_hermite_polynomial_he_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_i0e_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_i0e_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_i1_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_i1e_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_i1e_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_laguerre_polynomial_l_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_legendre_polynomial_p_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_legendre_polynomial_p_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_legendre_polynomial_p_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_log_ndtr_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_log_ndtr_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_modified_bessel_i1_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_modified_bessel_i1_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_modified_bessel_i1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_modified_bessel_i1_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_modified_bessel_k0_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_modified_bessel_k1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_modified_bessel_k1_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_ndtr_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_ndtri_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_ndtri_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_polygamma_special_polygamma_n_0_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_scaled_modified_bessel_k0_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_scaled_modified_bessel_k1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_shifted_chebyshev_polynomial_t_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_shifted_chebyshev_polynomial_t_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_shifted_chebyshev_polynomial_u_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_shifted_chebyshev_polynomial_u_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_shifted_chebyshev_polynomial_u_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_shifted_chebyshev_polynomial_v_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_shifted_chebyshev_polynomial_v_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_shifted_chebyshev_polynomial_w_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_shifted_chebyshev_polynomial_w_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_shifted_chebyshev_polynomial_w_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_shifted_chebyshev_polynomial_w_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_spherical_bessel_j0_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_spherical_bessel_j0_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_xlog1py_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_xlog1py_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_xlog1py_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_zeta_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_zeta_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_special_zeta_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_split_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_split_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_split_list_args_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_split_list_args_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_split_list_args_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_split_with_sizes_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_split_with_sizes_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_split_with_sizes_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_split_with_sizes_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_split_with_sizes_copy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_split_with_sizes_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_split_with_sizes_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sqrt_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_square_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_square_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_square_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_square_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_square_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_square_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_squeeze_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_squeeze_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_squeeze_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_squeeze_multiple_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_stack_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_stack_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_std_mean_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_std_mean_unbiased_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_std_mean_unbiased_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_std_unbiased_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_std_unbiased_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_std_unbiased_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sub_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sub_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sum_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sum_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sum_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sum_to_size_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_sum_to_size_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_svd_lowrank_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_t_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_t_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_t_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_t_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_take_along_dim_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_take_along_dim_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_take_along_dim_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_take_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_take_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_take_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_take_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_take_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_take_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tan_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tan_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tan_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tanh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tensor_split_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tensor_split_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tensordot_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tile_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tile_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_to_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_to_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_to_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_to_sparse_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_to_sparse_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_to_sparse_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_to_sparse_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_to_sparse_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_to_sparse_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_topk_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_topk_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_torch_ops_aten__efficient_attention_forward_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_torch_ops_aten__safe_softmax_default_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_torch_ops_aten__safe_softmax_default_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_trace_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_trace_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_trace_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_trace_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_transpose_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_transpose_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_transpose_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_transpose_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_trapezoid_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_trapezoid_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_trapezoid_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_trapezoid_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_trapezoid_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_triangular_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tril_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tril_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tril_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tril_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tril_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tril_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_tril_indices_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_triu_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_triu_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_true_divide_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_true_divide_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_trunc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_trunc_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unbind_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unbind_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unbind_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unflatten_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unflatten_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unfold_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unfold_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unfold_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unfold_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_uniform_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_uniform_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_uniform_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unique_consecutive_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unique_consecutive_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unique_cuda_uint32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unsafe_chunk_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unsafe_chunk_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unsafe_chunk_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unsafe_chunk_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unsafe_split_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unsafe_split_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unsafe_split_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unsafe_split_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unsafe_split_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unsafe_split_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unsqueeze_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unsqueeze_copy_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_unsqueeze_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_var_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_var_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_view_as_complex_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_view_as_complex_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_view_as_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_view_as_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_view_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_view_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_view_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_view_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_view_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_vsplit_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_vstack_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_vstack_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_where_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_where_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_where_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_where_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_where_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_xlogy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_zero__cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_zero__cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_zero__cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_zero__cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_zeros_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_zeros_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_zeros_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_zeros_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_inplace_zeros_like_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_H_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_T_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_T_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_T_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_T_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_T_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___getitem___cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___getitem___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___radd___cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___radd___cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___radd___cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rdiv___cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rdiv___cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rdiv___cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rmod___cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rmod___cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rmod___cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rmod___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rmul___cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rpow___cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rpow___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rsub___cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rsub___cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rsub___cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rsub___cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace___rsub___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__chunk_cat_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__chunk_cat_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_abs_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_abs_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_acos_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_acos_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_acos_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_add_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_add_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_add_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_addcdiv_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_addcdiv_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_addcdiv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_addcmul_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_addcmul_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_addcmul_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_addcmul_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_atan_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_atan_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_ceil_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_ceil_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_clamp_max_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_clamp_max_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_clamp_min_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_clamp_min_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_clamp_min_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_copy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_cos_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_cos_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_cosh_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_cosh_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_cosh_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_div_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_div_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_erf_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_erfc_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_erfc_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_erfc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_erfc_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_erfc_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_exp_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_exp_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_exp_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_expm1_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_expm1_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_floor_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_floor_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_floor_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_floor_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_floor_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_frac_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_lerp_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_lgamma_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_lgamma_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_lgamma_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_log10_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_log10_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_log10_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_log1p_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_log1p_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_log2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_log2_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_log2_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_log2_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_log2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_log2_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_log_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_maximum_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_maximum_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_minimum_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_minimum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_mul_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_neg_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_neg_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_neg_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_neg_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_norm_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_norm_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_pow_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_pow_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_pow_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_pow_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_reciprocal_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_reciprocal_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_reciprocal_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_round_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_round_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_round_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_round_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_round_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_sigmoid_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_sigmoid_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_sigmoid_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_sigmoid_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_sign_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_sign_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_sign_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_sin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_sin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_sin_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_sqrt_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_sqrt_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_sub_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_sub_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_tan_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_tan_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_tan_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_tan_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_tanh_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_tanh_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_tanh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_tanh_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_trunc_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_trunc_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_zero_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_zero_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_zero_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_zero_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__foreach_zero_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__segment_reduce_lengths_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__segment_reduce_offsets_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__softmax_backward_data_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__unsafe_masked_index_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__unsafe_masked_index_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__unsafe_masked_index_put_accumulate_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__unsafe_masked_index_put_accumulate_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__unsafe_masked_index_put_accumulate_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace__upsample_bilinear2d_aa_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_abs_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_abs_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_abs_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_abs_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_abs_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_abs_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_acos_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_acos_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_acosh_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_acosh_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_add_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_add_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_add_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_addbmm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_addcdiv_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_addcdiv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_addcmul_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_addmm_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_addmm_decomposed_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_addmm_decomposed_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_addmv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_addr_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_addr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_alias_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_alias_copy_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_alias_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_alias_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_alias_copy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_all_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_all_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_allclose_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_amax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_amax_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_amin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_aminmax_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_angle_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_angle_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_angle_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_any_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_any_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_arange_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_arange_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_argmax_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_argmax_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_argmin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_argmin_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_argmin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_argsort_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_argsort_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_argwhere_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_argwhere_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_argwhere_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_argwhere_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_argwhere_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_as_strided_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_as_strided_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_as_strided_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_as_strided_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_as_strided_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_as_strided_partial_views_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_as_strided_partial_views_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_as_strided_scatter_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_as_strided_scatter_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_as_strided_scatter_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_as_strided_scatter_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_asin_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_asin_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_asin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_asin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_asin_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_asinh_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_asinh_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_asinh_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atan2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atan_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atan_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atan_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atan_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atan_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atanh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atleast_1d_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atleast_1d_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atleast_2d_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atleast_2d_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atleast_2d_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atleast_2d_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atleast_2d_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atleast_3d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atleast_3d_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atleast_3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_atleast_3d_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_baddbmm_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bfloat16_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bincount_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bincount_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bincount_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bitwise_and_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bitwise_and_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bitwise_left_shift_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bitwise_or_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bitwise_right_shift_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bitwise_right_shift_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bitwise_right_shift_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bitwise_xor_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_block_diag_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bmm_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bmm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bool_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bool_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_broadcast_tensors_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_broadcast_tensors_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_broadcast_to_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_broadcast_to_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bucketize_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bucketize_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_bucketize_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_byte_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_byte_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cartesian_prod_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cartesian_prod_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cartesian_prod_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cat_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cat_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cat_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cauchy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cdouble_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ceil_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ceil_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cfloat_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cfloat_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_chalf_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_chalf_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_chalf_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_chalf_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cholesky_inverse_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cholesky_solve_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cholesky_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_chunk_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_clamp_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_clamp_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_clamp_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_clamp_max_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_clamp_max_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_clamp_min_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_clone_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_column_stack_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_combinations_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_complex_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_complex_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_conj_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_conj_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_conj_physical_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_constant_pad_nd_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_contiguous_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_contiguous_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_copysign_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_copysign_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_copysign_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_corrcoef_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_corrcoef_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_corrcoef_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_corrcoef_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cos_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cos_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cos_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cosh_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cosh_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cosh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_count_nonzero_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_count_nonzero_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_count_nonzero_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_count_nonzero_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_count_nonzero_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cov_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cov_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cov_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cov_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cross_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cross_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cross_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cummin_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cumprod_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cumsum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cumsum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cumulative_trapezoid_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cumulative_trapezoid_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_cumulative_trapezoid_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_deg2rad_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_deg2rad_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diag_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diag_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diag_embed_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diag_embed_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diag_embed_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diagflat_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diagflat_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diagonal_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diagonal_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diagonal_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diagonal_scatter_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diagonal_scatter_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diagonal_scatter_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diff_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diff_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diff_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_diff_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_digamma_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_dist_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_dist_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_dist_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_dist_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_div_floor_rounding_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_div_no_rounding_mode_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_div_no_rounding_mode_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_div_no_rounding_mode_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_div_no_rounding_mode_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_div_no_rounding_mode_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_div_trunc_rounding_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_div_trunc_rounding_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_double_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_dsplit_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_dsplit_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_einsum_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_einsum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_empty_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_empty_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_empty_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_empty_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_empty_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_empty_like_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_empty_permuted_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_empty_permuted_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_empty_permuted_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_empty_strided_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_empty_strided_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_empty_strided_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_empty_strided_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_eq_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_eq_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_eq_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_eq_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_eq_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_equal_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_erf_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_erf_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_erfc_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_erfc_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_erfinv_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_erfinv_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_erfinv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_erfinv_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_erfinv_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_exp2_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_exp2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_exp2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_exp2_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_exp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_exp_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_exp_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_exp_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_exp_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_expand_as_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_expand_as_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_expand_as_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_expand_as_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_expand_copy_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_expand_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_expm1_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_expm1_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_exponential_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_exponential_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_exponential_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_eye_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_eye_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_eye_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_fft2_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_fft_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_fft_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_fftn_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_fftn_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_fftn_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_fftshift_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_fftshift_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_fftshift_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_fftshift_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_hfft2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_hfft2_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_hfft_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_hfft_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_hfftn_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_hfftn_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_hfftn_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_hfftn_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_ifft2_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_ifft_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_ifft_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_ifftn_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_ifftn_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_ifftshift_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_ifftshift_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_ihfft2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_ihfft2_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_ihfft2_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_irfft2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_irfft_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_irfft_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_irfft_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_irfft_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_irfftn_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_rfft2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_rfft2_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_rfft2_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_rfftn_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fft_rfftn_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fill_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fill_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fill_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fill_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fill_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_flatten_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_flatten_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_flatten_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_flip_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_flip_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_flip_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_flip_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fliplr_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fliplr_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_flipud_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_float_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_float_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_float_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_float_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_float_power_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_floor_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_floor_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_floor_divide_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_floor_divide_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fmax_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fmax_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fmax_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fmax_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fmin_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fmin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fmod_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_fmod_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_frac_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_full_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_full_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_full_like_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_full_like_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_full_like_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_gather_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_gcd_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_geometric_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_geometric_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_geqrf_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_gradient_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_gradient_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_gradient_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_gradient_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_gt_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_half_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_half_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_heaviside_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_heaviside_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_histc_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_histc_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_hsplit_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_hsplit_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_hsplit_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_hsplit_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_hsplit_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_hsplit_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_hstack_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_hstack_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_hstack_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_hypot_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_i0_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_i0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_i0_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_igamma_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_add_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_add_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_add_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_put_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_reduce_amin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_reduce_amin_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_reduce_amin_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_reduce_mean_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_reduce_mean_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_reduce_mean_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_reduce_prod_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_select_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_select_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_index_select_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_inner_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_int_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_int_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_int_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_int_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isclose_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isclose_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isfinite_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isfinite_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isfinite_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isin_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isinf_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isinf_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isinf_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isnan_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isnan_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isneginf_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isneginf_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isposinf_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isposinf_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isreal_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isreal_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isreal_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_isreal_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_2inputs_2outputs_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_2inputs_2outputs_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_2inputs_2outputs_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_4inputs_with_extra_args_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_binary_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_binary_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_binary_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_binary_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_binary_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_binary_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_binary_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_binary_return_by_ref_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_unary_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_unary_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_jiterator_unary_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_kron_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_kthvalue_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_lcm_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ldexp_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ldexp_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ldexp_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ldexp_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ldexp_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_lerp_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_lerp_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_lerp_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_lerp_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_lerp_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_lgamma_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_lgamma_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_cond_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_cond_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_cross_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_det_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_diagonal_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_eig_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_eig_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_eigh_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_eigvals_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_householder_product_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_inv_ex_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_ldl_factor_ex_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_ldl_solve_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_ldl_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_lstsq_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_lstsq_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_lu_factor_ex_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_lu_factor_ex_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_matrix_norm_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_matrix_rank_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_matrix_rank_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_multi_dot_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_norm_subgradients_at_zero_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_pinv_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_pinv_singular_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_slogdet_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_solve_ex_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_solve_ex_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_solve_triangular_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_svd_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_svdvals_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_vander_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_vander_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_vander_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_vector_norm_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linalg_vector_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linspace_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_linspace_tensor_overload_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_log10_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_log10_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_log10_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_log10_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_log1p_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_log2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_log_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_log_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_log_softmax_with_dtype_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_log_softmax_with_dtype_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_log_softmax_with_dtype_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_log_softmax_with_dtype_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logaddexp_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logcumsumexp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logcumsumexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logdet_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logical_and_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logical_and_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logical_and_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logical_and_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logical_not_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logical_or_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logical_or_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logical_xor_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logical_xor_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logit_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logit_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logspace_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logspace_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logspace_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logspace_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logspace_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logspace_tensor_overload_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logspace_tensor_overload_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logspace_tensor_overload_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logsumexp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logsumexp_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_logsumexp_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_long_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_long_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_long_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_lu_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_lu_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_lu_unpack_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mT_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mT_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mT_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_amax_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_amax_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_amin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_argmax_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_argmin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_argmin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_argmin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_cumprod_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_cumsum_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_fill_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_fill_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_fill_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_logaddexp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_mean_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_mean_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_prod_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_prod_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_prod_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_prod_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_scatter_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_scatter_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_select_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_select_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_select_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_select_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_softmin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_std_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_sum_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_sum_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_masked_sum_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_max_binary_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_max_binary_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_max_binary_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_max_pool2d_with_indices_backward_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_max_pool2d_with_indices_backward_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_max_reduction_no_dim_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_max_reduction_with_dim_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_max_reduction_with_dim_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_max_reduction_with_dim_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_max_reduction_with_dim_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_max_reduction_with_dim_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_max_reduction_with_dim_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_maximum_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mean_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mean_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mean_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_median_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_meshgrid_list_of_tensors_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_meshgrid_list_of_tensors_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_meshgrid_variadic_tensors_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_min_binary_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_min_binary_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_min_binary_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_min_reduction_no_dim_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_min_reduction_no_dim_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_min_reduction_with_dim_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_minimum_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_minimum_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_minimum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mode_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mode_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mode_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mode_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_movedim_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_movedim_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_msort_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_msort_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mul_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mul_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mul_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mul_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mv_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mvlgamma_mvlgamma_p_1_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mvlgamma_mvlgamma_p_1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mvlgamma_mvlgamma_p_1_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mvlgamma_mvlgamma_p_3_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_mvlgamma_mvlgamma_p_3_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nan_to_num_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nan_to_num_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nan_to_num_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nan_to_num_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nanmean_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nansum_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_narrow_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_narrow_copy_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_narrow_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_narrow_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_narrow_copy_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_native_batch_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_native_batch_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_native_batch_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_native_layer_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ne_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ne_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ne_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ne_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ne_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ne_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_neg_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_new_empty_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_new_empty_strided_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_new_full_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_new_full_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_new_ones_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_new_ones_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_new_ones_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_new_ones_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_new_ones_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_new_zeros_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_new_zeros_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_adaptive_avg_pool1d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_adaptive_avg_pool1d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_adaptive_avg_pool1d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_adaptive_avg_pool3d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_adaptive_max_pool1d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_adaptive_max_pool3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_alpha_dropout_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_alpha_dropout_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_batch_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_batch_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_batch_norm_without_cudnn_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_binary_cross_entropy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_binary_cross_entropy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_binary_cross_entropy_with_logits_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_channel_shuffle_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_channel_shuffle_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_channel_shuffle_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_conv1d_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_conv1d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_conv2d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_conv3d_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_conv3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_conv_transpose3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_conv_transpose3d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_cosine_embedding_loss_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_cosine_embedding_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_cosine_embedding_loss_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_cross_entropy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_dropout3d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_dropout3d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_dropout_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_embedding_bag_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_embedding_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_feature_alpha_dropout_without_train_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_fractional_max_pool2d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_gaussian_nll_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_gelu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_glu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_glu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_grid_sample_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_group_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_group_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_hardswish_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_hardtanh_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_hardtanh_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_hinge_embedding_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_instance_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_instance_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_interpolate_area_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_interpolate_area_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_interpolate_bicubic_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_interpolate_nearest_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_interpolate_nearest_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_l1_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_l1_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_layer_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_layer_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_layer_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_linear_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_margin_ranking_loss_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_max_pool2d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_max_unpool1d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_max_unpool1d_grad_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_max_unpool2d_grad_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_max_unpool3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_mish_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_mse_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_multi_head_attention_forward_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_multi_margin_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_multilabel_margin_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_multilabel_soft_margin_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_normalize_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_normalize_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_normalize_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pad_circular_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pad_circular_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pad_circular_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pad_circular_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pad_constant_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pad_constant_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pad_reflect_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pad_reflect_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pad_reflect_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pad_replicate_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pad_replicate_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pad_replicate_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pad_replicate_negative_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pad_replicate_negative_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pairwise_distance_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pairwise_distance_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pairwise_distance_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pixel_shuffle_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pixel_shuffle_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pixel_shuffle_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pixel_shuffle_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_pixel_unshuffle_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_poisson_nll_loss_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_poisson_nll_loss_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_poisson_nll_loss_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_prelu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_prelu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_relu6_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_relu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_rms_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_rms_norm_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_selu_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_silu_complex_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_soft_margin_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_softmin_with_dtype_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_softmin_with_dtype_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_softplus_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_softsign_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_softsign_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_tanhshrink_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_tanhshrink_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_tanhshrink_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_threshold_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_threshold_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_triplet_margin_loss_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_triplet_margin_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_triplet_margin_loss_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_triplet_margin_with_distance_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_triplet_margin_with_distance_loss_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_unfold_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_unfold_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_upsample_bilinear_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nn_functional_upsample_nearest_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nonzero_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nonzero_static_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_nonzero_static_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_norm_fro_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_norm_fro_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_norm_inf_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_norm_nuc_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_normal_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ones_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ones_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ones_like_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ones_like_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ones_like_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ones_like_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ormqr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ormqr_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_outer_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_outer_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_outer_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_permute_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_permute_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_polygamma_polygamma_n_0_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_polygamma_polygamma_n_1_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_polygamma_polygamma_n_1_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_polygamma_polygamma_n_1_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_polygamma_polygamma_n_2_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_polygamma_polygamma_n_3_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_polygamma_polygamma_n_4_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_polygamma_polygamma_n_4_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_positive_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_positive_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_pow_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_pow_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_prod_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_prod_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_prod_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_put_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_put_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_qr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_quantile_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_rad2deg_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_rad2deg_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_rad2deg_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_rand_like_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_rand_like_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_randint_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_randint_like_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_randint_like_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_randn_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_randn_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_randn_like_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_randn_like_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ravel_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ravel_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_ravel_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_real_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_real_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_real_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_reciprocal_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_reciprocal_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_remainder_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_remainder_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_remainder_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_renorm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_repeat_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_repeat_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_repeat_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_repeat_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_repeat_interleave_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_repeat_interleave_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_repeat_interleave_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_repeat_interleave_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_repeat_interleave_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_repeat_interleave_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_reshape_as_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_reshape_as_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_reshape_as_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_reshape_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_reshape_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_resize__cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_resize__cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_resize__cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_resize_as__cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_resolve_conj_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_resolve_conj_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_resolve_conj_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_resolve_neg_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_roll_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_roll_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_roll_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_rot90_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_rot90_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_rot90_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_round_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_round_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_round_decimals_0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_round_decimals_3_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_round_decimals_3_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_round_decimals_3_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_rsub_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_rsub_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scalar_tensor_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scalar_tensor_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_add_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_add_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_add_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_add_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_reduce_amax_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_reduce_amin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_reduce_mean_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_reduce_mean_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_reduce_mean_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_reduce_prod_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_reduce_prod_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_reduce_prod_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_reduce_sum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_reduce_sum_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_scatter_reduce_sum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_searchsorted_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_select_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_select_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_select_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_select_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_select_scatter_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_select_scatter_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sgn_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sgn_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sgn_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_short_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_short_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_short_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_short_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sigmoid_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sign_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_signal_windows_exponential_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_signal_windows_general_hamming_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_signbit_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_signbit_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sin_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sinc_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sinc_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sinc_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sinc_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sinc_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sinc_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_slice_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_slice_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_slice_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_slice_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_slice_scatter_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_slice_scatter_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_slice_scatter_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_softmax_with_dtype_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_softmax_with_dtype_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_softmax_with_dtype_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sort_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sparse_mm_reduce_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_airy_ai_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_airy_ai_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_airy_ai_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_airy_ai_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_bessel_j0_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_bessel_j0_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_bessel_j1_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_bessel_j1_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_bessel_j1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_bessel_y0_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_bessel_y0_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_bessel_y0_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_bessel_y1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_bessel_y1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_chebyshev_polynomial_t_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_chebyshev_polynomial_t_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_chebyshev_polynomial_u_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_chebyshev_polynomial_u_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_chebyshev_polynomial_v_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_chebyshev_polynomial_v_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_chebyshev_polynomial_v_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_chebyshev_polynomial_w_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_chebyshev_polynomial_w_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_entr_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_entr_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_entr_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_entr_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_erfcx_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_erfcx_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_hermite_polynomial_he_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_hermite_polynomial_he_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_hermite_polynomial_he_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_i0e_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_i0e_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_i1e_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_i1e_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_laguerre_polynomial_l_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_legendre_polynomial_p_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_legendre_polynomial_p_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_log_ndtr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_log_ndtr_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_log_ndtr_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_modified_bessel_i0_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_modified_bessel_i1_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_modified_bessel_i1_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_modified_bessel_i1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_modified_bessel_i1_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_modified_bessel_k0_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_modified_bessel_k0_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_ndtr_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_ndtr_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_ndtri_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_polygamma_special_polygamma_n_0_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_scaled_modified_bessel_k0_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_scaled_modified_bessel_k0_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_scaled_modified_bessel_k1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_scaled_modified_bessel_k1_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_shifted_chebyshev_polynomial_t_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_shifted_chebyshev_polynomial_u_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_shifted_chebyshev_polynomial_v_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_shifted_chebyshev_polynomial_w_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_shifted_chebyshev_polynomial_w_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_spherical_bessel_j0_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_spherical_bessel_j0_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_spherical_bessel_j0_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_xlog1py_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_xlog1py_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_xlog1py_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_zeta_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_zeta_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_special_zeta_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_split_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_split_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_split_list_args_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_split_with_sizes_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_split_with_sizes_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_split_with_sizes_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_split_with_sizes_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_split_with_sizes_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_split_with_sizes_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_split_with_sizes_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sqrt_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_square_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_square_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_square_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_square_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_square_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_squeeze_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_stack_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_stack_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_stack_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_std_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_std_mean_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_std_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_std_mean_unbiased_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_std_mean_unbiased_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_std_mean_unbiased_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_std_mean_unbiased_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sub_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sum_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sum_to_size_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_sum_to_size_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_svd_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_t_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_t_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_t_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_t_copy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_t_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_t_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_t_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_t_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_take_along_dim_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_take_along_dim_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_take_along_dim_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_take_along_dim_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_take_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_take_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_take_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_take_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_tan_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_tan_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_tan_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_tan_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_tanh_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_tanh_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_tanh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_tensor_split_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_tensor_split_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_tensordot_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_tile_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_to_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_to_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_to_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_topk_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_topk_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_torch_ops_aten__safe_softmax_default_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_trace_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_transpose_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_transpose_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_transpose_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_transpose_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_trapezoid_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_trapezoid_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_trapezoid_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_trapz_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_tril_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_triu_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_true_divide_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_trunc_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_trunc_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unbind_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unbind_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unbind_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unbind_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unflatten_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unflatten_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unflatten_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unfold_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unfold_copy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unfold_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unfold_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_uniform_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unique_consecutive_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unique_consecutive_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unique_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unique_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unique_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unique_cuda_uint16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unravel_index_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unravel_index_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unsafe_chunk_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unsafe_split_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unsafe_split_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unsqueeze_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unsqueeze_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_unsqueeze_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_var_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_var_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_var_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_var_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_var_mean_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_var_mean_unbiased_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_var_mean_unbiased_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_vdot_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_view_as_complex_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_view_as_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_view_as_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_view_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_view_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_view_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_view_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_vsplit_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_vsplit_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_vsplit_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_vstack_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_vstack_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_vstack_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_where_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_where_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_where_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_xlogy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_xlogy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_xlogy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_xlogy_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_xlogy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_zero__cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_zero__cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_zeros_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_zeros_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_zeros_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_zeros_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_zeros_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_meta_outplace_zeros_like_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_H_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_H_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_T_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_T_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___getitem___cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___getitem___cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___getitem___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___rdiv___cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___rdiv___cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___rdiv___cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___rmatmul___cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___rmatmul___cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___rmod___cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___rmod___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___rmul___cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___rmul___cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___ror___cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___rpow___cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___rsub___cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___rsub___cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace___rsub___cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__batch_norm_with_update_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__chunk_cat_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__chunk_cat_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__chunk_cat_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__chunk_cat_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_abs_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_abs_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_add_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_add_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_add_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_add_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_addcdiv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_addcdiv_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_addcmul_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_addcmul_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_addcmul_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_addcmul_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_addcmul_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_asin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_asin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_asin_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_asin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_asin_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_atan_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_atan_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_atan_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_atan_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_ceil_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_ceil_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_clamp_max_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_clamp_max_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_clamp_max_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_clamp_min_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_cos_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_cos_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_cosh_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_cosh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_div_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_div_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_erf_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_erf_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_erf_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_erfc_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_exp_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_exp_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_expm1_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_expm1_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_expm1_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_floor_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_floor_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_floor_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_frac_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_lerp_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_lgamma_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_log10_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_log10_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_log10_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_log1p_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_log1p_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_log2_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_log2_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_log2_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_log2_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_log2_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_log_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_log_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_max_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_max_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_max_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_max_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_minimum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_minimum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_mul_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_mul_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_mul_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_mul_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_neg_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_neg_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_neg_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_norm_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_norm_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_norm_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_pow_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_pow_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_pow_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_pow_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_pow_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_reciprocal_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_reciprocal_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_round_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_round_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_sigmoid_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_sigmoid_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_sign_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_sign_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_sin_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_sin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_sin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_sqrt_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_sqrt_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_sub_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_sub_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_sub_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_sub_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_tanh_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_tanh_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_tanh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_tanh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_trunc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_zero_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_zero_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_zero_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_zero_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_zero_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__foreach_zero_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__segment_reduce_offsets_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__softmax_backward_data_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__unsafe_masked_index_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__unsafe_masked_index_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__unsafe_masked_index_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__unsafe_masked_index_put_accumulate_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace__unsafe_masked_index_put_accumulate_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_abs_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_abs_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_abs_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_abs_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_acos_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_acos_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_acos_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_acos_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_acosh_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_add_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_add_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_addbmm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_addcmul_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_addmm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_addmm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_addr_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_addr_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_addr_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_alias_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_alias_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_alias_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides___rpow___cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides__foreach_atan_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides__foreach_frac_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides__foreach_log10_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides__foreach_maximum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides__foreach_sigmoid_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides__foreach_tan_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides__segment_reduce_lengths_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_addmm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_addmv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_addr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_aminmax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_argmax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_atan2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_baddbmm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_bitwise_or_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_bitwise_xor_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_block_diag_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_bmm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_ceil_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_cholesky_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_complex_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_cummax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_diag_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_diagonal_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_digamma_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_dist_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_empty_strided_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_equal_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_erfinv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_expand_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_exponential_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_eye_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_fft_fft_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_fft_ifftshift_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_fft_ihfft_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_fft_irfft_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_flatten_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_floor_divide_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_fmin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_frexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_full_like_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_gather_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_ge_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_geometric_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_grid_sampler_2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_gt_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_hsplit_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_imag_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_index_fill_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_index_reduce_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_inner_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_int_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_item_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_jiterator_2inputs_2outputs_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_lgamma_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_linalg_det_singular_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_linalg_diagonal_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_linalg_eigh_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_linalg_eigvals_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_linalg_ldl_factor_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_linalg_lstsq_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_linalg_matrix_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_linalg_matrix_power_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_linalg_matrix_rank_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_linalg_svd_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_linalg_tensorinv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_linspace_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_log_normal_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_logical_and_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_logical_not_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_logical_xor_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_lu_unpack_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_masked_argmin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_masked_median_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_max_pool2d_with_indices_backward_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_min_binary_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_min_reduction_with_dim_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_mode_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_movedim_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_msort_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nanmean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nanmedian_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_narrow_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_neg_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_new_full_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_new_ones_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_adaptive_avg_pool3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_avg_pool2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_celu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_conv1d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_conv3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_conv_transpose1d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_feature_alpha_dropout_with_train_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_fractional_max_pool2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_fractional_max_pool3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_interpolate_area_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_interpolate_linear_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_interpolate_nearest-exact_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_l1_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_logsigmoid_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_max_unpool3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_mse_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_normalize_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_relu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_silu_complex_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_silu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_softsign_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nn_functional_upsample_nearest_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_nonzero_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_norm_nuc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_ones_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_ones_like_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_polygamma_polygamma_n_3_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_rand_like_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_randn_like_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_round_decimals_3_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_scalar_tensor_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_signal_windows_bartlett_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_signal_windows_gaussian_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_signal_windows_nuttall_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_special_i0e_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_special_i1e_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_special_modified_bessel_i0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_special_ndtr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_split_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_split_list_args_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_split_with_sizes_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_squeeze_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_svd_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_svd_lowrank_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_take_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_tanh_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_tile_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_triangular_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_unflatten_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_unravel_index_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_var_unbiased_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_view_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_view_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_xlogy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_all_strides_zeros_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_amax_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_amax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_amax_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_amax_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_amin_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_aminmax_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_aminmax_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_angle_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_angle_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_any_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_any_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_any_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_any_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_arange_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_arange_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_argmax_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_argmax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_argmax_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_argmin_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_argsort_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_argwhere_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_argwhere_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_argwhere_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_argwhere_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_argwhere_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_as_strided_copy_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_as_strided_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_as_strided_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_as_strided_copy_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_as_strided_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_as_strided_copy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_as_strided_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_as_strided_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_as_strided_partial_views_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_as_strided_partial_views_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_as_strided_partial_views_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_as_strided_scatter_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_asin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_asin_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_asin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_asin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_asinh_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_asinh_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_asinh_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_atan2_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_atan_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_atan_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_atan_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_atanh_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_atleast_1d_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_atleast_2d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_atleast_2d_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_atleast_3d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_atleast_3d_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_atleast_3d_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_baddbmm_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_bernoulli_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_bfloat16_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_bfloat16_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_bfloat16_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_bitwise_and_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_bitwise_and_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_bitwise_left_shift_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_bitwise_not_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_bitwise_right_shift_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_bitwise_xor_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_bitwise_xor_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_block_diag_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_block_diag_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_broadcast_tensors_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_broadcast_tensors_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_broadcast_to_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_broadcast_to_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_broadcast_to_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_broadcast_to_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_byte_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_byte_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cartesian_prod_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cartesian_prod_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cat_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cauchy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cdouble_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_ceil_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cfloat_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cfloat_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_chalf_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_chalf_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_chalf_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_chalf_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_char_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_char_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_char_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cholesky_inverse_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cholesky_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_chunk_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_chunk_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_clamp_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_clamp_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_clamp_max_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_clamp_min_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_clamp_min_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_clamp_min_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_clamp_min_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_clone_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_clone_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_clone_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_clone_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_column_stack_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_column_stack_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_column_stack_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_combinations_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_combinations_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_combinations_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_combinations_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_conj_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_conj_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_conj_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_conj_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_conj_physical_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_conj_physical_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_constant_pad_nd_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_constant_pad_nd_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_constant_pad_nd_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_contiguous_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_copysign_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_copysign_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_copysign_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_corrcoef_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_corrcoef_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_corrcoef_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cos_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cos_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cosh_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cosh_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cosh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_count_nonzero_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cross_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cross_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cross_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cross_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cross_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cummax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cummax_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cumprod_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cumprod_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cumsum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cumulative_trapezoid_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cumulative_trapezoid_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_cumulative_trapezoid_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_deg2rad_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_deg2rad_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_deg2rad_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_deg2rad_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diag_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diag_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diag_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diag_embed_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diag_embed_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diag_embed_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diagflat_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diagflat_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diagflat_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diagonal_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diagonal_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diagonal_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diagonal_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diagonal_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diagonal_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diagonal_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diagonal_scatter_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diagonal_scatter_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diagonal_scatter_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diff_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diff_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diff_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diff_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diff_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_diff_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_digamma_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_dist_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_dist_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_div_floor_rounding_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_div_floor_rounding_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_div_no_rounding_mode_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_div_no_rounding_mode_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_div_trunc_rounding_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_div_trunc_rounding_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_div_trunc_rounding_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_double_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_double_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_double_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_double_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_double_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_dsplit_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_dsplit_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_dsplit_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_dstack_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_dstack_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_dstack_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_einsum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_einsum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_empty_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_empty_like_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_empty_like_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_empty_permuted_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_empty_permuted_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_empty_strided_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_empty_strided_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_empty_strided_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_eq_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_eq_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_equal_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_equal_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_erf_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_erf_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_erfc_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_erfinv_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_erfinv_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_exp2_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_exp_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_exp_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_exp_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_exp_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_expand_as_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_expand_as_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_expand_as_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_expand_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_expand_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_expand_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_expand_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_expand_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_expand_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_expand_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_expm1_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_expm1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_expm1_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_expm1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_exponential_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_fft2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_fft2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_fft_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_fft_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_fft_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_fftn_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_fftn_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_fftn_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_fftn_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_fftshift_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_fftshift_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_fftshift_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_hfft2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_hfft2_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_hfft_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_hfftn_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_hfftn_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_hfftn_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_hfftn_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ifft2_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ifft2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ifft2_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ifft_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ifftn_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ifftn_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ifftn_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ifftshift_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ifftshift_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ifftshift_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ifftshift_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ifftshift_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ihfft_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ihfft_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ihfftn_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ihfftn_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ihfftn_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_ihfftn_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_irfft2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_irfft2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_irfft_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_irfftn_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_rfft2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fft_rfft2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fill_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_flatten_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_flip_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fliplr_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fliplr_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fliplr_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_float_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_float_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_float_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_float_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_floor_divide_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fmax_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_fmod_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_frac_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_frac_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_frexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_full_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_full_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_full_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_full_like_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_full_like_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_full_like_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_gather_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_gather_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_gather_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_gather_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_ge_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_ge_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_ge_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_geometric_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_geqrf_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_geqrf_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_gradient_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_grid_sampler_2d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_half_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_half_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_histc_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_hsplit_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_hstack_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_i0_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_i0_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_i0_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_imag_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_add_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_add_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_add_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_add_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_add_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_put_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_put_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_reduce_amax_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_reduce_amax_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_reduce_amin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_reduce_mean_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_reduce_prod_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_select_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_select_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_select_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_select_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_index_select_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_int_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isclose_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isclose_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isclose_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isclose_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isclose_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isfinite_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isfinite_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isin_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isinf_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isinf_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isnan_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isnan_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isnan_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isnan_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isneginf_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isneginf_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isposinf_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isposinf_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isreal_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_isreal_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_istft_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_item_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_item_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_jiterator_2inputs_2outputs_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_jiterator_2inputs_2outputs_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_jiterator_2inputs_2outputs_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_jiterator_4inputs_with_extra_args_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_jiterator_4inputs_with_extra_args_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_jiterator_4inputs_with_extra_args_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_jiterator_4inputs_with_extra_args_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_jiterator_binary_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_jiterator_binary_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_jiterator_binary_return_by_ref_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_jiterator_unary_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_jiterator_unary_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_jiterator_unary_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_jiterator_unary_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_kron_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_kron_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_kron_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_kron_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_kthvalue_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_lcm_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_ldexp_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_le_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_le_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_lerp_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_cholesky_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_cholesky_ex_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_cholesky_ex_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_cholesky_ex_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_cond_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_cross_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_cross_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_cross_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_cross_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_det_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_det_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_det_singular_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_diagonal_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_diagonal_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_diagonal_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_eig_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_eigh_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_eigvalsh_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_inv_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_inv_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_ldl_factor_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_ldl_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_lstsq_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_lstsq_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_lu_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_lu_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_lu_factor_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_lu_factor_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_lu_factor_ex_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_lu_solve_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_matrix_norm_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_matrix_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_matrix_rank_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_matrix_rank_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_matrix_rank_hermitian_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_matrix_rank_hermitian_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_multi_dot_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_pinv_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_pinv_hermitian_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_pinv_hermitian_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_slogdet_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_slogdet_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_solve_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_solve_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_solve_triangular_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_svdvals_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_vander_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_vecdot_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linalg_vecdot_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linspace_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linspace_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linspace_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linspace_tensor_overload_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linspace_tensor_overload_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_linspace_tensor_overload_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_log10_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_log10_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_log1p_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_log2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_log2_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_log_normal_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_log_normal_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_log_softmax_with_dtype_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_log_softmax_with_dtype_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_log_softmax_with_dtype_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_log_softmax_with_dtype_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logaddexp_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logdet_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logdet_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logical_and_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logical_and_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logical_not_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logical_not_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logical_or_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logical_or_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logical_xor_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logical_xor_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logical_xor_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logit_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logit_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logit_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logspace_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logspace_tensor_overload_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logspace_tensor_overload_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logsumexp_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logsumexp_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logsumexp_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_logsumexp_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_long_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_long_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_long_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_lt_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_lu_unpack_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_lu_unpack_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mH_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mH_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mH_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mT_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mT_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_amin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_amin_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_amin_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_argmax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_argmax_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_argmin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_argmin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_cumprod_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_cumprod_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_cumsum_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_fill_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_fill_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_fill_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_fill_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_fill_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_log_softmax_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_logaddexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_logsumexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_logsumexp_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_mean_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_median_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_prod_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_prod_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_prod_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_scatter_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_scatter_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_select_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_softmax_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_softmin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_std_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_std_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_std_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_std_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_sum_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_sum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_sum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_masked_var_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_matrix_exp_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_matrix_exp_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_matrix_exp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_max_binary_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_max_binary_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_max_reduction_no_dim_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_max_reduction_no_dim_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_max_reduction_no_dim_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_max_reduction_no_dim_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_max_reduction_with_dim_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_max_reduction_with_dim_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_maximum_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mean_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_median_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_meshgrid_list_of_tensors_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_meshgrid_list_of_tensors_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_meshgrid_list_of_tensors_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_meshgrid_list_of_tensors_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_meshgrid_list_of_tensors_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_meshgrid_list_of_tensors_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_meshgrid_variadic_tensors_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_meshgrid_variadic_tensors_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_min_binary_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_min_binary_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_min_binary_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_min_reduction_no_dim_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_min_reduction_no_dim_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_min_reduction_no_dim_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_min_reduction_with_dim_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_min_reduction_with_dim_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_min_reduction_with_dim_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_min_reduction_with_dim_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mm_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mode_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_movedim_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_movedim_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_movedim_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_movedim_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_movedim_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_msort_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_msort_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mul_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mul_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mul_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mul_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_multinomial_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mv_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mvlgamma_mvlgamma_p_1_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mvlgamma_mvlgamma_p_3_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mvlgamma_mvlgamma_p_3_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_mvlgamma_mvlgamma_p_3_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nan_to_num_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nan_to_num_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nanmean_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nansum_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_narrow_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_narrow_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_native_batch_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_native_layer_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_ne_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_neg_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_neg_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_neg_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_neg_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_empty_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_empty_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_empty_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_empty_strided_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_empty_strided_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_empty_strided_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_empty_strided_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_empty_strided_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_full_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_full_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_full_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_full_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_ones_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_ones_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_ones_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_ones_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_zeros_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_zeros_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_new_zeros_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_adaptive_avg_pool2d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_adaptive_max_pool3d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_alpha_dropout_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_batch_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_batch_norm_without_cudnn_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_binary_cross_entropy_with_logits_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_channel_shuffle_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_channel_shuffle_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_channel_shuffle_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_conv1d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_conv2d_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_conv3d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_conv_transpose1d_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_conv_transpose1d_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_conv_transpose1d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_conv_transpose1d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_conv_transpose2d_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_conv_transpose2d_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_conv_transpose3d_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_cosine_embedding_loss_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_cosine_embedding_loss_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_dropout2d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_dropout2d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_dropout3d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_elu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_elu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_embedding_bag_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_embedding_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_feature_alpha_dropout_with_train_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_fractional_max_pool3d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_gaussian_nll_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_gaussian_nll_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_group_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_hardsigmoid_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_hardsigmoid_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_hardtanh_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_instance_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_instance_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_interpolate_area_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_interpolate_bicubic_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_interpolate_bicubic_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_interpolate_linear_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_interpolate_trilinear_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_kl_div_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_kl_div_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_kl_div_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_l1_loss_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_l1_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_layer_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_leaky_relu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_linear_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_linear_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_margin_ranking_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_margin_ranking_loss_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_max_pool1d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_max_pool1d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_max_pool1d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_max_pool2d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_max_pool3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_max_pool3d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_max_unpool1d_grad_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_max_unpool2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_max_unpool3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_max_unpool3d_grad_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_max_unpool3d_grad_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_mish_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_mish_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_multi_head_attention_forward_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_multilabel_margin_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_nll_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_normalize_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_normalize_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_circular_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_circular_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_circular_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_circular_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_circular_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_circular_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_constant_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_constant_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_constant_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_constant_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_reflect_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_reflect_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_reflect_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_replicate_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_replicate_negative_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pad_replicate_negative_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pairwise_distance_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pairwise_distance_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pixel_shuffle_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pixel_shuffle_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pixel_shuffle_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pixel_shuffle_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_pixel_unshuffle_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_poisson_nll_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_poisson_nll_loss_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_poisson_nll_loss_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_poisson_nll_loss_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_relu6_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_relu6_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_relu6_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_relu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_relu_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_relu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_relu_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_scaled_dot_product_attention_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_selu_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_soft_margin_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_softmin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_softmin_with_dtype_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_softplus_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_tanhshrink_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_tanhshrink_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_tanhshrink_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_threshold_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_threshold_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_threshold_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_triplet_margin_loss_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_triplet_margin_with_distance_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nn_functional_unfold_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nonzero_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nonzero_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nonzero_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nonzero_static_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nonzero_static_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_nonzero_static_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_norm_fro_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_norm_inf_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_normal_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_normal_in_place_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_normal_in_place_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_ones_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_ones_like_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_ormqr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_outer_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_outer_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_outer_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_outer_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_pca_lowrank_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_permute_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_permute_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_pinverse_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_polar_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_polar_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_polygamma_polygamma_n_0_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_polygamma_polygamma_n_3_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_polygamma_polygamma_n_3_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_polygamma_polygamma_n_3_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_polygamma_polygamma_n_4_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_polygamma_polygamma_n_4_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_polygamma_polygamma_n_4_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_positive_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_pow_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_pow_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_pow_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_pow_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_prod_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_put_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_put_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_put_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_put_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_put_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_qr_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_qr_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_rad2deg_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_rad2deg_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_rad2deg_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_rand_like_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_randint_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_randint_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_randint_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_randint_like_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_randn_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_randn_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_randn_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_ravel_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_real_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_real_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_reciprocal_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_remainder_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_remainder_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_renorm_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_renorm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_renorm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_renorm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_repeat_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_repeat_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_repeat_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_repeat_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_repeat_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_repeat_interleave_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_repeat_interleave_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_reshape_as_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_reshape_as_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_reshape_as_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_reshape_as_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_reshape_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_reshape_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_reshape_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_reshape_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_resize_as__cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_resize_as__cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_resolve_conj_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_resolve_conj_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_resolve_conj_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_resolve_conj_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_resolve_neg_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_resolve_neg_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_resolve_neg_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_roll_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_rot90_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_rot90_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_rot90_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_rot90_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_round_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_round_decimals_0_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_round_decimals_0_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_round_decimals_neg_3_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_rsqrt_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_rsqrt_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_rsqrt_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scalar_tensor_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_add_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_add_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_add_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_reduce_amax_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_reduce_amax_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_reduce_amin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_reduce_amin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_reduce_prod_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_reduce_prod_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_reduce_prod_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_reduce_sum_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_reduce_sum_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_reduce_sum_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_scatter_reduce_sum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_searchsorted_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_searchsorted_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_searchsorted_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_select_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sgn_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sgn_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sgn_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_short_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_short_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sigmoid_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sigmoid_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sign_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sign_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sign_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_signal_windows_cosine_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_signal_windows_exponential_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_signal_windows_gaussian_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_signal_windows_general_cosine_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_signbit_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_signbit_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sinc_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sinh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_slice_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_slice_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_slice_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_slice_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_slice_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_slice_scatter_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_slice_scatter_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_softmax_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_softmax_with_dtype_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_softmax_with_dtype_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_softmax_with_dtype_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_airy_ai_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_bessel_j0_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_bessel_j0_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_bessel_j1_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_bessel_j1_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_bessel_j1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_bessel_j1_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_bessel_j1_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_bessel_y0_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_chebyshev_polynomial_t_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_chebyshev_polynomial_u_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_chebyshev_polynomial_v_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_chebyshev_polynomial_w_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_chebyshev_polynomial_w_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_entr_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_entr_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_erfcx_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_erfcx_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_hermite_polynomial_h_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_hermite_polynomial_he_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_hermite_polynomial_he_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_i0e_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_i0e_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_i1e_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_laguerre_polynomial_l_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_laguerre_polynomial_l_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_laguerre_polynomial_l_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_legendre_polynomial_p_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_log_ndtr_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_modified_bessel_i0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_modified_bessel_i0_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_modified_bessel_i1_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_modified_bessel_k0_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_modified_bessel_k1_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_ndtr_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_ndtri_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_polygamma_special_polygamma_n_0_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_scaled_modified_bessel_k0_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_scaled_modified_bessel_k0_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_scaled_modified_bessel_k1_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_scaled_modified_bessel_k1_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_scaled_modified_bessel_k1_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_shifted_chebyshev_polynomial_t_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_shifted_chebyshev_polynomial_t_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_shifted_chebyshev_polynomial_u_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_shifted_chebyshev_polynomial_u_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_shifted_chebyshev_polynomial_u_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_shifted_chebyshev_polynomial_v_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_shifted_chebyshev_polynomial_w_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_shifted_chebyshev_polynomial_w_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_shifted_chebyshev_polynomial_w_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_spherical_bessel_j0_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_spherical_bessel_j0_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_spherical_bessel_j0_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_xlog1py_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_xlog1py_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_xlog1py_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_xlog1py_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_xlog1py_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_xlog1py_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_zeta_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_zeta_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_zeta_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_zeta_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_special_zeta_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_split_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_split_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_split_list_args_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_split_with_sizes_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_split_with_sizes_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_split_with_sizes_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_split_with_sizes_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_split_with_sizes_copy_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_split_with_sizes_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_split_with_sizes_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sqrt_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_squeeze_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_squeeze_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_squeeze_multiple_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_squeeze_multiple_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_stack_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_stack_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_stack_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_stack_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_stack_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_stack_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_stack_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_std_mean_unbiased_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_std_unbiased_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_std_unbiased_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sub_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sub_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sub_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sum_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sum_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sum_to_size_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sum_to_size_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_sum_to_size_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_svd_lowrank_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_svd_lowrank_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_t_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_t_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_t_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_t_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_t_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_take_along_dim_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_take_along_dim_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_take_along_dim_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_take_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_tan_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_tan_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_tan_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_tanh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_tensor_split_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_tensor_split_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_tensor_split_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_tensor_split_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_tile_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_tile_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_tile_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_to_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_to_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_to_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_to_sparse_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_to_sparse_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_to_sparse_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_topk_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_topk_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_torch_ops_aten__safe_softmax_default_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_torch_ops_aten__safe_softmax_default_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_torch_ops_aten__safe_softmax_default_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_trace_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_trace_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_trace_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_trace_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_transpose_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_transpose_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_transpose_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_transpose_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_trapezoid_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_trapezoid_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_trapz_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_trapz_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_trapz_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_triangular_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_tril_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_tril_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_tril_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_triu_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_triu_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_true_divide_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_true_divide_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_trunc_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_trunc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_trunc_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unbind_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unbind_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unbind_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unflatten_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unflatten_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unfold_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unfold_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unfold_copy_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unfold_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unfold_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unfold_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unfold_copy_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unfold_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unfold_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unfold_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_uniform_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unique_consecutive_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unique_consecutive_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unique_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unique_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unique_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unravel_index_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unravel_index_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unsafe_chunk_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unsafe_chunk_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unsafe_split_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unsafe_split_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unsafe_split_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unsafe_split_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unsqueeze_copy_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unsqueeze_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_unsqueeze_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_var_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_var_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_var_mean_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_var_unbiased_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_var_unbiased_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_var_unbiased_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_vdot_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_vdot_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_view_as_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_view_as_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_view_as_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_view_as_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_view_as_real_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_view_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_view_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_view_copy_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_view_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_vsplit_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_vsplit_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_vsplit_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_vstack_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_where_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_xlogy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_xlogy_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_zero__cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_zeros_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_zeros_like_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_zeros_like_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_zeros_like_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_zeros_like_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_inplace_zeros_like_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_H_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_H_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_H_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_H_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_T_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_T_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___getitem___cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___getitem___cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___getitem___cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___getitem___cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___radd___cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___radd___cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rand___cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rand___cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rand___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rdiv___cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rdiv___cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rdiv___cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rdiv___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rmatmul___cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rmod___cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rmod___cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rmod___cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rmod___cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rmod___cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rmod___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rmul___cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rmul___cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___ror___cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rpow___cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rpow___cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rsub___cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rsub___cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rsub___cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rxor___cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace___rxor___cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__batch_norm_with_update_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__chunk_cat_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__chunk_cat_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_abs_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_abs_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_abs_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_abs_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_acos_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_acos_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_acos_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_add_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_add_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_add_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_add_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_addcdiv_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_addcmul_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_asin_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_asin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_asin_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_atan_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_atan_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_ceil_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_ceil_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_ceil_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_clamp_max_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_clamp_max_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_clamp_min_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_clamp_min_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_cos_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_cos_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_cos_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_cosh_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_cosh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_cosh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_div_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_div_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_erf_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_erf_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_erf_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_erf_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_erfc_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_exp_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_exp_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_exp_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_expm1_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_floor_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_floor_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_floor_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_floor_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_frac_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_frac_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_frac_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_frac_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_frac_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_frac_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_lerp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_lerp_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_lerp_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_lerp_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_lgamma_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_log10_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_log10_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_log10_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_log2_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_log2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_log_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_max_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_max_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_max_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_maximum_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_maximum_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_maximum_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_minimum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_minimum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_mul_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_neg_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_neg_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_neg_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_neg_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_neg_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_norm_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_norm_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_pow_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_reciprocal_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_reciprocal_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_round_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_round_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_round_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sigmoid_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sign_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sign_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sign_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sign_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sign_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sin_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sin_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sinh_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sqrt_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sqrt_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sqrt_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sqrt_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sub_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sub_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_sub_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_tan_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_tan_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_tan_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_tan_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_tanh_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_tanh_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_trunc_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_trunc_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_trunc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_trunc_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_zero_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_zero_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__foreach_zero_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__segment_reduce_offsets_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__softmax_backward_data_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__unsafe_masked_index_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__unsafe_masked_index_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__unsafe_masked_index_put_accumulate_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__unsafe_masked_index_put_accumulate_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__unsafe_masked_index_put_accumulate_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__upsample_bilinear2d_aa_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace__upsample_bilinear2d_aa_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_abs_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_abs_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_acos_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_acos_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_acos_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_acos_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_acosh_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_acosh_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_add_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_addcmul_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_addmm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_addmm_decomposed_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_addmm_decomposed_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_addmv_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_addmv_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_addr_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_alias_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_alias_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_T_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides___rand___cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides___rmod___cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides___rmul___cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides__foreach_abs_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides__foreach_addcdiv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides__foreach_asin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides__foreach_clamp_min_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides__foreach_cos_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides__foreach_cosh_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides__foreach_log2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides__foreach_sign_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides__foreach_sub_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides__foreach_zero_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides__native_batch_norm_legit_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides__softmax_backward_data_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides__upsample_bilinear2d_aa_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_add_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_addr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_argmin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_argwhere_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_asin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_atleast_2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_baddbmm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_bernoulli_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_bitwise_not_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_bitwise_right_shift_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_broadcast_shapes_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_cdist_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_cdouble_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_conj_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_cummin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_diag_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_empty_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_eq_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_exp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_expm1_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_fft_hfft2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_fft_hfft_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_flip_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_fmin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_fmod_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_gradient_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_grid_sampler_2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_imag_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_index_reduce_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_index_reduce_prod_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_isfinite_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_isinf_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_isneginf_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_jiterator_4inputs_with_extra_args_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_le_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_linalg_cond_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_linalg_eigh_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_linalg_lstsq_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_linalg_matrix_power_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_linalg_matrix_rank_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_linalg_pinv_hermitian_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_linalg_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_linalg_tensorsolve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_log2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_log_softmax_with_dtype_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_logcumsumexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_logical_not_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_lt_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_lu_unpack_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_mT_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_masked_cumprod_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_masked_cumsum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_masked_log_softmax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_masked_median_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_masked_prod_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_masked_softmax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_masked_sum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_matrix_exp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_max_binary_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_max_reduction_no_dim_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_max_reduction_with_dim_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_min_reduction_with_dim_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nanmean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_narrow_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_new_empty_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_new_full_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_adaptive_avg_pool3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_binary_cross_entropy_with_logits_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_conv3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_dropout2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_dropout3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_embedding_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_interpolate_area_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_interpolate_bilinear_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_layer_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_logsigmoid_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_margin_ranking_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_max_unpool2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_max_unpool2d_grad_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_multi_head_attention_forward_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_relu6_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_scaled_dot_product_attention_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_softmin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_softsign_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_tanhshrink_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_threshold_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_nn_functional_upsample_nearest_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_norm_fro_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_norm_nuc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_outer_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_pow_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_put_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_randn_like_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_remainder_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_repeat_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_resize_as__cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_rot90_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_round_decimals_3_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_scalar_tensor_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_scatter_reduce_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_signal_windows_blackman_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_signal_windows_hann_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_slice_scatter_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_softmax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_special_airy_ai_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_special_bessel_y0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_special_chebyshev_polynomial_w_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_special_log_ndtr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_special_modified_bessel_k1_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_special_shifted_chebyshev_polynomial_t_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_std_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_std_mean_unbiased_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_std_unbiased_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_t_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_take_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_tan_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_tensor_split_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_to_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_to_sparse_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_triangular_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_tril_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_unbind_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_unique_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_unsqueeze_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_view_as_complex_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_where_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_zero__cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_all_strides_zeros_like_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_allclose_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_allclose_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_amax_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_amax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_amin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_aminmax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_aminmax_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_angle_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_any_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_any_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_arange_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_arange_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_arange_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_arange_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_argmax_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_argmax_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_argmin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_argmin_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_argmin_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_argsort_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_argsort_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_argsort_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_argsort_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_argwhere_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_argwhere_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_argwhere_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_as_strided_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_as_strided_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_as_strided_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_as_strided_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_as_strided_partial_views_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_as_strided_partial_views_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_as_strided_partial_views_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_as_strided_scatter_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_asin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_asinh_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_asinh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_atan_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_atan_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_atanh_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_atanh_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_atanh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_atleast_1d_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_atleast_2d_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_atleast_2d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_atleast_3d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_atleast_3d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_atleast_3d_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_atleast_3d_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_baddbmm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bfloat16_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bitwise_and_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bitwise_and_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bitwise_and_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bitwise_left_shift_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bitwise_or_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bitwise_or_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bitwise_or_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bitwise_right_shift_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bitwise_right_shift_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bitwise_xor_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_block_diag_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bmm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bmm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bool_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_bool_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_broadcast_tensors_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_broadcast_tensors_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_broadcast_tensors_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_broadcast_tensors_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_broadcast_to_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_broadcast_to_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_broadcast_to_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_broadcast_to_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_broadcast_to_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_byte_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cartesian_prod_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cartesian_prod_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cat_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cat_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cat_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cat_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cat_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cdouble_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ceil_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cfloat_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cfloat_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cfloat_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_chalf_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_chalf_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_char_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_char_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cholesky_inverse_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cholesky_solve_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_chunk_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_chunk_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_clamp_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_clamp_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_clamp_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_clamp_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_clamp_max_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_clone_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_clone_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_column_stack_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_column_stack_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_column_stack_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_combinations_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_combinations_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_complex_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_conj_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_conj_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_conj_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_conj_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_conj_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_conj_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_conj_physical_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_conj_physical_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_conj_physical_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_constant_pad_nd_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_constant_pad_nd_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_constant_pad_nd_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_constant_pad_nd_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_contiguous_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_contiguous_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_contiguous_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_copysign_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_corrcoef_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_corrcoef_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_corrcoef_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cos_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cosh_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cosh_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cosh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cosh_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_count_nonzero_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_count_nonzero_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_count_nonzero_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_count_nonzero_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cov_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cov_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cross_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cross_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cummax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cummax_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cummin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cummin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cumprod_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cumsum_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cumulative_trapezoid_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cumulative_trapezoid_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_cumulative_trapezoid_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_deg2rad_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diag_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diag_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diag_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diag_embed_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diag_embed_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diag_embed_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diagflat_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diagflat_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diagflat_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diagonal_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diagonal_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diagonal_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diagonal_copy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diagonal_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diagonal_scatter_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diagonal_scatter_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diagonal_scatter_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diagonal_scatter_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diff_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diff_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_diff_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_digamma_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_digamma_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_dist_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_div_floor_rounding_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_div_floor_rounding_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_div_no_rounding_mode_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_div_no_rounding_mode_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_div_no_rounding_mode_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_div_no_rounding_mode_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_div_trunc_rounding_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_div_trunc_rounding_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_div_trunc_rounding_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_dot_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_dot_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_dot_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_double_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_dsplit_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_dstack_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_dstack_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_dstack_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_dstack_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_dstack_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_einsum_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_empty_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_empty_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_empty_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_empty_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_empty_like_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_empty_permuted_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_empty_permuted_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_empty_strided_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_empty_strided_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_empty_strided_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_empty_strided_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_eq_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_eq_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_equal_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_equal_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_equal_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_equal_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_erf_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_erf_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_erfc_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_erfc_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_erfinv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_exp2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_exp2_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_exp2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_exp2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_exp2_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_exp_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_expand_as_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_expand_as_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_expand_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_expand_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_expand_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_expand_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_expand_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_expand_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_expm1_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_expm1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_exponential_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_eye_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_fft2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_fft2_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_fft2_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_fft_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_fftn_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_fftshift_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_fftshift_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_fftshift_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_hfft2_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_hfft2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_hfft_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_hfft_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_hfft_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_hfftn_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_ifft2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_ifft2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_ifft2_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_ifft_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_ifft_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_ifft_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_ifftn_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_ifftshift_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_ihfft2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_ihfft_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_ihfft_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_ihfftn_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_ihfftn_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_ihfftn_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_irfft2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_irfft2_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_irfft_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_irfft_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_irfft_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_irfft_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_irfftn_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_rfft2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fft_rfft2_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fill_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fill_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fill_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fill_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_flatten_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_flatten_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fliplr_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fliplr_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fliplr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fliplr_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fliplr_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fliplr_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_flipud_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_flipud_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_flipud_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_float_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_float_power_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_floor_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_floor_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_floor_divide_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fmax_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fmin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fmin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fmin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_fmin_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_frac_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_frac_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_frexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_full_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_full_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_full_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_full_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_full_like_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_full_like_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_gather_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_gather_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_gather_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_gather_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ge_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ge_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_geometric_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_geometric_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_gradient_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_gradient_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_gradient_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_grid_sampler_2d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_gt_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_half_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_half_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_half_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_half_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_half_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_half_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_heaviside_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_heaviside_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_hsplit_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_hsplit_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_hsplit_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_hstack_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_hstack_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_i0_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_i0_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_i0_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_imag_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_add_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_add_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_add_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_add_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_copy_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_copy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_put_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_put_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_put_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_reduce_amax_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_reduce_amin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_reduce_amin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_reduce_amin_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_reduce_amin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_reduce_mean_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_reduce_mean_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_reduce_mean_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_reduce_prod_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_reduce_prod_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_select_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_select_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_select_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_index_select_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_int_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_int_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_int_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isclose_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isclose_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isfinite_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isfinite_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isinf_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isinf_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isinf_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isnan_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isnan_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isnan_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isnan_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isneginf_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isneginf_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isneginf_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isposinf_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isposinf_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_isreal_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_item_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_jiterator_2inputs_2outputs_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_jiterator_4inputs_with_extra_args_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_jiterator_4inputs_with_extra_args_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_jiterator_4inputs_with_extra_args_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_jiterator_4inputs_with_extra_args_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_jiterator_binary_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_jiterator_binary_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_jiterator_binary_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_jiterator_binary_return_by_ref_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_jiterator_unary_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_jiterator_unary_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_jiterator_unary_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_kron_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_kron_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_kron_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_kron_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_kthvalue_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_kthvalue_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_lcm_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_lcm_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ldexp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ldexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ldexp_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ldexp_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_le_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_le_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_le_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_lerp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_lgamma_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_lgamma_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_cholesky_ex_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_cond_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_cross_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_cross_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_det_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_det_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_diagonal_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_diagonal_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_eig_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_eig_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_eigvals_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_eigvals_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_eigvals_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_householder_product_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_householder_product_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_inv_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_inv_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_ldl_factor_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_ldl_solve_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_lu_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_lu_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_lu_factor_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_lu_factor_ex_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_lu_solve_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_matrix_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_matrix_rank_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_matrix_rank_hermitian_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_norm_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_norm_subgradients_at_zero_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_norm_subgradients_at_zero_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_pinv_hermitian_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_pinv_singular_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_qr_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_solve_triangular_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_vander_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_vander_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_vander_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_vecdot_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_vecdot_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linalg_vecdot_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linspace_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linspace_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linspace_tensor_overload_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linspace_tensor_overload_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_linspace_tensor_overload_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_log10_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_log10_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_log10_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_log1p_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_log1p_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_log1p_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_log2_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_log_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_log_softmax_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_log_softmax_with_dtype_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_log_softmax_with_dtype_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_log_softmax_with_dtype_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logaddexp2_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logcumsumexp_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logcumsumexp_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logcumsumexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logdet_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logdet_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logical_not_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logical_not_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logical_not_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logical_or_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logical_xor_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logical_xor_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logit_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logit_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logit_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logspace_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logspace_tensor_overload_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logspace_tensor_overload_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logspace_tensor_overload_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logsumexp_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_logsumexp_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_long_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_long_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_long_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_lt_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_lt_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_lu_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mH_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mT_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mT_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mT_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mT_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_amax_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_amin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_amin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_amin_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_argmax_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_argmax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_argmax_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_argmin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_argmin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_cumprod_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_cumsum_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_cumsum_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_fill_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_fill_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_fill_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_fill_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_logaddexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_logaddexp_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_logsumexp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_logsumexp_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_logsumexp_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_mean_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_mean_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_median_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_normalize_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_normalize_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_normalize_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_normalize_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_prod_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_prod_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_prod_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_prod_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_scatter_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_scatter_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_scatter_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_select_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_select_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_softmax_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_softmin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_std_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_sum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_var_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_var_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_masked_var_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_matrix_exp_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_max_binary_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_max_binary_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_max_binary_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_max_pool2d_with_indices_backward_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_max_pool2d_with_indices_backward_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_max_reduction_no_dim_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_max_reduction_no_dim_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_max_reduction_no_dim_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_maximum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mean_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mean_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_median_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_meshgrid_list_of_tensors_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_meshgrid_list_of_tensors_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_meshgrid_variadic_tensors_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_min_binary_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_min_binary_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_min_reduction_no_dim_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_min_reduction_no_dim_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_min_reduction_with_dim_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_min_reduction_with_dim_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_min_reduction_with_dim_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mode_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mode_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mode_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mode_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mode_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_movedim_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_movedim_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_msort_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mul_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mul_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mul_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_multinomial_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mvlgamma_mvlgamma_p_1_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mvlgamma_mvlgamma_p_1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mvlgamma_mvlgamma_p_1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mvlgamma_mvlgamma_p_3_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mvlgamma_mvlgamma_p_3_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_mvlgamma_mvlgamma_p_5_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nan_to_num_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nan_to_num_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nanmean_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nanmedian_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nansum_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nansum_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nansum_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nansum_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_narrow_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_narrow_copy_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_narrow_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_narrow_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_narrow_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_narrow_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_native_batch_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_native_dropout_backward_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_native_dropout_backward_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_native_dropout_backward_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_neg_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_neg_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_new_empty_strided_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_new_empty_strided_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_new_empty_strided_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_new_full_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_new_full_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_new_full_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_new_full_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_new_ones_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_new_ones_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_new_ones_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_new_zeros_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_new_zeros_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nextafter_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_adaptive_avg_pool1d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_adaptive_avg_pool3d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_adaptive_max_pool1d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_adaptive_max_pool2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_adaptive_max_pool3d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_avg_pool2d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_batch_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_batch_norm_without_cudnn_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_bilinear_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_bilinear_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_binary_cross_entropy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_binary_cross_entropy_with_logits_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_channel_shuffle_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_channel_shuffle_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_channel_shuffle_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_channel_shuffle_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_conv3d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_conv3d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_conv3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_conv3d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_conv_transpose1d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_cosine_embedding_loss_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_dropout3d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_dropout_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_elu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_elu_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_embedding_bag_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_embedding_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_embedding_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_feature_alpha_dropout_without_train_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_fractional_max_pool2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_gaussian_nll_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_gelu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_grid_sample_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_grid_sample_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_group_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_hardswish_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_hardtanh_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_hardtanh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_huber_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_instance_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_instance_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_interpolate_area_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_interpolate_area_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_interpolate_area_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_interpolate_linear_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_interpolate_nearest-exact_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_interpolate_nearest-exact_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_interpolate_nearest_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_interpolate_nearest_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_interpolate_trilinear_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_interpolate_trilinear_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_interpolate_trilinear_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_l1_loss_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_leaky_relu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_leaky_relu_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_linear_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_linear_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_local_response_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_logsigmoid_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_logsigmoid_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_margin_ranking_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_margin_ranking_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_margin_ranking_loss_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_max_pool1d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_max_pool2d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_max_pool2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_max_unpool1d_grad_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_max_unpool2d_grad_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_mish_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_mish_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_mse_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_multi_head_attention_forward_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_multi_margin_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_multilabel_margin_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_normalize_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_normalize_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_normalize_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pad_circular_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pad_circular_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pad_circular_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pad_circular_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pad_circular_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pad_constant_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pad_constant_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pad_replicate_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pad_replicate_negative_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pad_replicate_negative_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pairwise_distance_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pixel_unshuffle_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pixel_unshuffle_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pixel_unshuffle_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_pixel_unshuffle_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_poisson_nll_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_poisson_nll_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_poisson_nll_loss_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_poisson_nll_loss_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_prelu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_prelu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_relu6_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_relu6_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_relu6_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_relu6_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_relu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_relu_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_relu_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_rms_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_rms_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_rrelu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_rrelu_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_selu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_silu_complex_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_silu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_silu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_smooth_l1_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_softmin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_softmin_with_dtype_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_softmin_with_dtype_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_softshrink_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_softsign_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_softsign_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_softsign_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_softsign_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_tanhshrink_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_tanhshrink_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_tanhshrink_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_tanhshrink_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_threshold_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_threshold_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_threshold_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_triplet_margin_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_triplet_margin_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_triplet_margin_loss_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_triplet_margin_loss_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_triplet_margin_with_distance_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_triplet_margin_with_distance_loss_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nn_functional_unfold_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nonzero_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nonzero_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nonzero_static_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nonzero_static_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nonzero_static_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nonzero_static_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_nonzero_static_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_norm_fro_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_norm_inf_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_norm_nuc_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_normal_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_normal_number_mean_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ones_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ones_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ones_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ones_like_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ones_like_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ones_like_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ones_like_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ones_like_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_outer_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_outer_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_outer_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_permute_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_permute_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_permute_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_permute_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_polygamma_polygamma_n_0_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_polygamma_polygamma_n_0_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_polygamma_polygamma_n_0_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_polygamma_polygamma_n_1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_polygamma_polygamma_n_1_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_polygamma_polygamma_n_2_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_polygamma_polygamma_n_2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_polygamma_polygamma_n_3_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_polygamma_polygamma_n_3_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_polygamma_polygamma_n_3_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_polygamma_polygamma_n_4_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_polygamma_polygamma_n_4_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_positive_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_positive_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_positive_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_pow_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_pow_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_pow_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_pow_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_prod_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_prod_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_put_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_put_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_put_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_put_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_rad2deg_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_rand_like_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_rand_like_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_randint_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_randint_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_randint_like_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_randint_like_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_randn_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_randn_like_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_randn_like_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ravel_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ravel_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_ravel_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_real_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_real_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_reciprocal_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_reciprocal_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_reciprocal_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_reciprocal_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_remainder_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_remainder_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_remainder_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_remainder_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_repeat_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_repeat_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_repeat_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_repeat_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_repeat_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_resize__cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_resize__cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_resize__cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_resize__cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_resize__cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_resize_as__cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_resize_as__cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_resolve_conj_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_resolve_neg_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_resolve_neg_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_resolve_neg_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_roll_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_roll_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_roll_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_rot90_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_rot90_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_round_decimals_0_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_round_decimals_0_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_round_decimals_3_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_round_decimals_neg_3_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_rsqrt_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_rsqrt_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_rsqrt_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_rsub_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_rsub_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scalar_tensor_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scalar_tensor_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scalar_tensor_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scalar_tensor_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scalar_tensor_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_add_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_add_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_add_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_add_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_reduce_amax_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_reduce_amax_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_reduce_amax_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_reduce_amin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_reduce_amin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_reduce_mean_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_reduce_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_reduce_prod_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_reduce_sum_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_reduce_sum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_scatter_reduce_sum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_searchsorted_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_select_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_select_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_select_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_select_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_select_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_select_scatter_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_select_scatter_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_select_scatter_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_select_scatter_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_select_scatter_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sgn_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sgn_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sgn_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sgn_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_short_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_short_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_short_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sigmoid_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sigmoid_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sign_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sign_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_signal_windows_bartlett_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_signal_windows_blackman_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_signal_windows_cosine_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_signal_windows_nuttall_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_signbit_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_signbit_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_signbit_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sin_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sinc_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sinc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sinc_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sinc_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sinh_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sinh_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sinh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sinh_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_slice_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_slice_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_slice_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_slice_scatter_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_slice_scatter_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_softmax_with_dtype_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_softmax_with_dtype_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_softmax_with_dtype_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_softmax_with_dtype_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sort_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sort_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sparse_mm_reduce_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_bessel_j0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_bessel_j0_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_bessel_j1_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_bessel_j1_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_bessel_y0_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_chebyshev_polynomial_t_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_chebyshev_polynomial_t_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_chebyshev_polynomial_t_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_chebyshev_polynomial_w_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_chebyshev_polynomial_w_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_entr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_entr_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_erfcx_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_erfcx_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_i0e_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_i0e_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_i1e_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_i1e_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_i1e_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_laguerre_polynomial_l_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_laguerre_polynomial_l_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_laguerre_polynomial_l_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_legendre_polynomial_p_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_legendre_polynomial_p_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_legendre_polynomial_p_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_log_ndtr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_log_ndtr_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_modified_bessel_i0_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_modified_bessel_i0_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_modified_bessel_i1_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_modified_bessel_i1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_modified_bessel_i1_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_modified_bessel_i1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_modified_bessel_k0_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_ndtr_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_ndtri_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_polygamma_special_polygamma_n_0_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_polygamma_special_polygamma_n_0_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_polygamma_special_polygamma_n_0_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_scaled_modified_bessel_k0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_scaled_modified_bessel_k1_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_scaled_modified_bessel_k1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_scaled_modified_bessel_k1_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_scaled_modified_bessel_k1_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_shifted_chebyshev_polynomial_t_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_shifted_chebyshev_polynomial_u_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_shifted_chebyshev_polynomial_u_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_shifted_chebyshev_polynomial_u_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_shifted_chebyshev_polynomial_v_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_shifted_chebyshev_polynomial_v_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_shifted_chebyshev_polynomial_w_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_shifted_chebyshev_polynomial_w_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_shifted_chebyshev_polynomial_w_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_spherical_bessel_j0_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_spherical_bessel_j0_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_xlog1py_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_xlog1py_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_xlog1py_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_zeta_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_special_zeta_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_split_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_split_list_args_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_split_with_sizes_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_split_with_sizes_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_split_with_sizes_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sqrt_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sqrt_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_square_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_square_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_square_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_squeeze_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_squeeze_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_squeeze_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_squeeze_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_squeeze_multiple_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_stack_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_stack_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_stack_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_std_mean_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_std_mean_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_stft_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sub_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sub_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sub_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sum_to_size_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sum_to_size_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sum_to_size_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sum_to_size_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_sum_to_size_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_t_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_t_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_t_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_take_along_dim_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_take_along_dim_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_take_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_take_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_take_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_take_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_take_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_tan_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_tan_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_tanh_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_tanh_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_tanh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_tensor_split_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_tensor_split_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_tensor_split_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_tensordot_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_tile_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_tile_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_to_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_to_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_to_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_to_sparse_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_to_sparse_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_to_sparse_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_to_sparse_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_topk_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_torch__scaled_mm_cuda_float8_e4m3fn, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_torch_ops_aten__safe_softmax_default_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_torch_ops_aten__safe_softmax_default_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_trace_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_trace_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_transpose_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_transpose_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_transpose_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_transpose_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_transpose_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_trapezoid_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_trapezoid_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_trapz_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_trapz_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_triangular_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_tril_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_tril_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_tril_indices_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_triu_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_true_divide_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_true_divide_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_true_divide_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_true_divide_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unbind_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unbind_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unbind_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unbind_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unbind_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unflatten_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unflatten_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unflatten_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unflatten_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unfold_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unfold_copy_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unfold_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unfold_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unfold_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unfold_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unfold_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unfold_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_uniform_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unique_consecutive_cuda_bool, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unique_consecutive_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unique_consecutive_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unique_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unique_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unravel_index_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unravel_index_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unsafe_chunk_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unsafe_chunk_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unsafe_chunk_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unsafe_chunk_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unsafe_split_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unsafe_split_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unsqueeze_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unsqueeze_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unsqueeze_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unsqueeze_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unsqueeze_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_unsqueeze_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_var_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_var_mean_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_var_mean_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_var_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_var_mean_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_var_unbiased_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_vdot_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_vdot_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_view_as_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_view_as_cuda_float64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_view_as_real_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_view_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_view_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_view_copy_cuda_int32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_view_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_view_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_view_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_vsplit_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_vsplit_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_vsplit_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_vstack_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_vstack_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_vstack_cuda_float16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_where_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_where_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_where_cuda_int16, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_xlogy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_zeros_cuda_int64, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_zeros_like_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_zeros_like_cuda_float32, test/test_meta.py::TestMetaCUDA::test_dispatch_symbolic_meta_outplace_zeros_like_cuda_int64, test/test_meta.py::TestMetaCUDA::test_empty_quantized_cuda, test/test_meta.py::TestMetaCUDA::test_layer_norm_backward_output_mask0_cuda, test/test_meta.py::TestMetaCUDA::test_layer_norm_backward_output_mask4_cuda, test/test_meta.py::TestMetaCUDA::test_layer_norm_backward_output_mask5_cuda, test/test_meta.py::TestMetaCUDA::test_meta__fused_moving_avg_obs_fq_helper_cuda, test/test_meta.py::TestMetaCUDA::test_meta_inplace_H_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_H_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_T_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace___radd___cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace___radd___cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace___radd___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace___rdiv___cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace___rdiv___cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace___rdiv___cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace___rmatmul___cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace___rmod___cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace___rmod___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace___rmul___cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace___ror___cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace___rpow___cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace___rpow___cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__batch_norm_with_update_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__chunk_cat_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__chunk_cat_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__chunk_cat_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__chunk_cat_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_acos_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_acos_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_acos_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_add_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_add_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_addcdiv_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_addcdiv_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_addcdiv_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_addcdiv_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_addcmul_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_addcmul_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_asin_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_asin_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_atan_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_atan_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_ceil_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_ceil_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_ceil_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_ceil_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_clamp_max_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_clamp_max_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_clamp_max_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_clamp_max_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_clamp_min_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_clamp_min_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_cos_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_erf_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_erf_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_erf_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_exp_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_expm1_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_expm1_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_expm1_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_floor_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_floor_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_frac_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_lerp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_lerp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_lgamma_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_lgamma_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_log10_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_log10_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_log1p_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_log1p_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_log1p_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_log2_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_log_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_log_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_max_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_maximum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_minimum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_minimum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_minimum_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_mul_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_mul_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_mul_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_neg_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_norm_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_pow_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_pow_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_pow_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_reciprocal_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_reciprocal_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_round_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_round_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_round_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sigmoid_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sigmoid_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sigmoid_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sigmoid_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sigmoid_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sign_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sign_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sign_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sinh_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sinh_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sinh_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sqrt_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sqrt_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_sub_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_tan_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_tanh_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_tanh_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_tanh_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_tanh_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_trunc_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_trunc_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_zero_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__foreach_zero_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__native_batch_norm_legit_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__segment_reduce_lengths_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__segment_reduce_offsets_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__softmax_backward_data_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__unsafe_masked_index_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace__unsafe_masked_index_put_accumulate_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace__unsafe_masked_index_put_accumulate_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace__unsafe_masked_index_put_accumulate_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace__upsample_bilinear2d_aa_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_abs_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_acos_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_acos_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_acosh_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_acosh_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_acosh_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_add_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_add_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_add_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_add_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addbmm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addcdiv_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addcdiv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addcmul_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addcmul_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addcmul_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addcmul_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addmm_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addmm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addmm_decomposed_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addr_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addr_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addr_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addr_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_addr_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_alias_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_alias_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_all_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_amax_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_amin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_aminmax_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_angle_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_angle_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_any_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_any_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_any_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_argmax_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_argmin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_argmin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_argsort_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_argsort_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_argwhere_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_as_strided_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_as_strided_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_as_strided_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_as_strided_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_as_strided_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_as_strided_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_as_strided_partial_views_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_as_strided_partial_views_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_as_strided_partial_views_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_as_strided_partial_views_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_as_strided_scatter_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_as_strided_scatter_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_as_strided_scatter_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_as_strided_scatter_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_asin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_asin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_asinh_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_asinh_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atan2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atan_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atanh_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atanh_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atanh_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atanh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atanh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atleast_1d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atleast_1d_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atleast_1d_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atleast_1d_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atleast_2d_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atleast_2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atleast_2d_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atleast_2d_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atleast_3d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atleast_3d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atleast_3d_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_atleast_3d_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_baddbmm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bernoulli_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bfloat16_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bfloat16_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bfloat16_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bfloat16_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bitwise_and_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bitwise_or_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bitwise_right_shift_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bitwise_right_shift_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bitwise_xor_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bitwise_xor_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bmm_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bmm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bool_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bool_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bool_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_broadcast_tensors_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_broadcast_tensors_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_broadcast_to_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_broadcast_to_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_broadcast_to_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_broadcast_to_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bucketize_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_bucketize_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_byte_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cartesian_prod_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cat_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cat_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cdouble_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cdouble_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cdouble_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ceil_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ceil_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cfloat_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cfloat_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cfloat_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cfloat_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_chalf_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_char_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_char_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cholesky_solve_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_chunk_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_chunk_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_chunk_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_chunk_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_chunk_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_clamp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_clamp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_clamp_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_clamp_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_clamp_max_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_clamp_max_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_clamp_min_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_clamp_min_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_clone_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_clone_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_clone_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_clone_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_column_stack_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_column_stack_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_combinations_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_combinations_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_combinations_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_conj_physical_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_conj_physical_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_constant_pad_nd_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_copysign_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_copysign_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_corrcoef_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_corrcoef_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_corrcoef_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_count_nonzero_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_count_nonzero_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cov_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cov_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cov_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cross_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cross_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cummax_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cummax_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cummin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cummin_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cumsum_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cumsum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cumsum_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_cumsum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_deg2rad_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_diag_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_diag_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_diag_embed_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_diag_embed_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_diag_embed_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_diag_embed_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_diagflat_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_diagflat_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_diagonal_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_diagonal_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_diagonal_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_diagonal_scatter_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_diagonal_scatter_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_diff_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_dist_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_dist_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_dist_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_div_floor_rounding_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_div_floor_rounding_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_div_no_rounding_mode_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_div_no_rounding_mode_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_div_no_rounding_mode_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_div_no_rounding_mode_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_div_no_rounding_mode_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_div_trunc_rounding_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_div_trunc_rounding_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_dot_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_dot_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_dot_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_double_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_dsplit_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_dsplit_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_dsplit_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_dsplit_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_dstack_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_dstack_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_einsum_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_einsum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_empty_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_empty_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_empty_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_empty_like_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_empty_like_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_empty_like_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_empty_permuted_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_empty_permuted_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_empty_strided_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_eq_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_equal_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_equal_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_erf_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_erf_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_erfc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_erfinv_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_erfinv_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_exp2_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_exp2_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_exp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_exp_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expand_as_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expand_as_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expand_as_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expand_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expand_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expand_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expand_copy_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expand_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expand_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expand_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expand_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expm1_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expm1_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expm1_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_expm1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_exponential_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_exponential_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_eye_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_eye_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_eye_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_eye_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fft2_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fft2_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fft2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fft_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fft_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fft_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fftn_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fftn_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fftn_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fftn_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fftn_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fftn_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fftn_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fftshift_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_fftshift_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_hfft2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_hfft2_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_hfft_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_hfft_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_hfft_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_hfftn_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_hfftn_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_hfftn_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_hfftn_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_ifft2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_ifft2_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_ifft2_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_ifftshift_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_ifftshift_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_ihfft2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_ihfft2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_ihfft2_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_ihfft2_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_ihfftn_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_ihfftn_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_ihfftn_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_irfft2_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_irfft2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_irfft2_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_irfft_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_irfft_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_irfft_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_irfftn_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_irfftn_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_rfft2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fft_rfft_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fill_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fill_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_flatten_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_flatten_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_flatten_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_flatten_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_flip_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fliplr_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fliplr_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fliplr_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fliplr_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_flipud_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_float_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_float_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_float_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_float_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_float_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_float_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_float_power_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_float_power_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_float_power_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_floor_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_floor_divide_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fmax_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fmax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fmax_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fmax_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fmin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fmin_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_fmin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_full_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_full_like_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_full_like_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ge_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ge_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ge_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_geqrf_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_geqrf_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_geqrf_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_grid_sampler_2d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_grid_sampler_2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_gt_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_gt_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_half_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_half_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_heaviside_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_histc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_histc_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_histc_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_histc_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_histc_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_histc_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_hsplit_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_hsplit_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_hsplit_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_hstack_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_hstack_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_hstack_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_hstack_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_hypot_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_hypot_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_hypot_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_i0_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_igamma_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_igammac_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_igammac_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_index_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_index_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_index_fill_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_index_fill_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_index_fill_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_index_fill_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_index_fill_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_index_reduce_amax_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_index_reduce_amin_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_index_reduce_mean_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_index_reduce_mean_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_index_reduce_prod_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_index_select_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_index_select_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_inner_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_inner_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_int_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_int_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_int_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_int_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_int_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isfinite_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isfinite_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isinf_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isinf_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isinf_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isnan_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isnan_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isnan_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isnan_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isneginf_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isposinf_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isposinf_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isreal_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isreal_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isreal_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_isreal_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_istft_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_item_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_item_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_item_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_2inputs_2outputs_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_2inputs_2outputs_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_2inputs_2outputs_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_4inputs_with_extra_args_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_4inputs_with_extra_args_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_4inputs_with_extra_args_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_4inputs_with_extra_args_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_binary_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_binary_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_binary_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_binary_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_binary_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_binary_return_by_ref_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_unary_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_unary_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_jiterator_unary_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_kron_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_kthvalue_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_lcm_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_lcm_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ldexp_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ldexp_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ldexp_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_le_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_le_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_lerp_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_lgamma_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_lgamma_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_cholesky_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_cholesky_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_cond_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_cond_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_cross_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_cross_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_det_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_diagonal_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_diagonal_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_diagonal_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_diagonal_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_eig_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_eigvals_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_inv_ex_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_ldl_factor_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_ldl_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_lu_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_lu_factor_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_lu_factor_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_lu_factor_ex_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_lu_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_lu_solve_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_matrix_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_matrix_norm_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_matrix_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_matrix_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_matrix_rank_hermitian_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_matrix_rank_hermitian_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_multi_dot_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_multi_dot_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_norm_subgradients_at_zero_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_norm_subgradients_at_zero_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_norm_subgradients_at_zero_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_pinv_hermitian_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_pinv_singular_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_pinv_singular_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_slogdet_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_svd_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_svdvals_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_vander_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_vander_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_vecdot_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linalg_vector_norm_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_linspace_tensor_overload_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_log10_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_log10_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_log10_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_log1p_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_log2_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_log_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_log_normal_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_log_softmax_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_log_softmax_with_dtype_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_log_softmax_with_dtype_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_log_softmax_with_dtype_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_log_softmax_with_dtype_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logaddexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logaddexp_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logical_and_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logical_and_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logical_and_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logical_and_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logical_not_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logical_or_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logical_or_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logical_or_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logical_or_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logical_xor_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logical_xor_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logical_xor_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logit_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logit_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logspace_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logspace_tensor_overload_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logsumexp_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logsumexp_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logsumexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_logsumexp_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_long_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_long_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_long_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_long_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_lt_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_lt_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mH_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mH_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mT_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mT_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mT_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_amax_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_amax_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_amax_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_amin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_argmax_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_argmin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_argmin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_argmin_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_cumprod_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_cumprod_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_cumprod_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_cumprod_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_cumsum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_cumsum_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_fill_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_fill_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_fill_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_logaddexp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_logaddexp_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_logsumexp_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_logsumexp_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_mean_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_mean_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_normalize_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_normalize_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_prod_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_prod_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_prod_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_prod_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_scatter_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_scatter_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_select_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_softmax_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_softmax_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_softmin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_std_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_std_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_sum_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_sum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_var_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_var_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_masked_var_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_matmul_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_matmul_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_matrix_exp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_max_binary_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_max_binary_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_max_reduction_no_dim_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_max_reduction_no_dim_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_max_reduction_no_dim_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_max_reduction_with_dim_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_max_reduction_with_dim_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_max_reduction_with_dim_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_max_reduction_with_dim_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_maximum_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_maximum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_median_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_meshgrid_list_of_tensors_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_meshgrid_list_of_tensors_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_meshgrid_list_of_tensors_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_meshgrid_list_of_tensors_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_meshgrid_variadic_tensors_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_meshgrid_variadic_tensors_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_min_binary_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_min_binary_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_min_binary_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_min_reduction_no_dim_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_min_reduction_with_dim_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_minimum_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_minimum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mm_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mode_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mode_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_movedim_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_movedim_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_msort_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_msort_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mul_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mv_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mv_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mvlgamma_mvlgamma_p_1_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mvlgamma_mvlgamma_p_3_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mvlgamma_mvlgamma_p_3_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mvlgamma_mvlgamma_p_3_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mvlgamma_mvlgamma_p_3_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mvlgamma_mvlgamma_p_3_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mvlgamma_mvlgamma_p_5_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_mvlgamma_mvlgamma_p_5_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nan_to_num_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nanmean_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nanmean_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nanmedian_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nanmedian_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nanquantile_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_narrow_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_narrow_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_narrow_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_narrow_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_narrow_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_narrow_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_narrow_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_native_batch_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_native_layer_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_native_layer_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_native_layer_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_neg_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_neg_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_new_empty_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_new_empty_strided_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_new_empty_strided_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_new_empty_strided_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_new_full_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_new_full_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_new_full_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_new_ones_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_new_ones_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_new_ones_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_new_zeros_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_new_zeros_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_adaptive_avg_pool1d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_adaptive_avg_pool1d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_adaptive_avg_pool2d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_adaptive_avg_pool3d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_adaptive_max_pool3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_avg_pool1d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_avg_pool3d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_batch_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_batch_norm_without_cudnn_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_bilinear_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_binary_cross_entropy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_binary_cross_entropy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_binary_cross_entropy_with_logits_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_binary_cross_entropy_with_logits_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_celu_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_celu_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_channel_shuffle_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_channel_shuffle_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_channel_shuffle_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_conv1d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_conv2d_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_conv2d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_conv3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_conv3d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_conv_transpose1d_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_cosine_embedding_loss_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_cosine_embedding_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_cosine_embedding_loss_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_cosine_similarity_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_cosine_similarity_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_cosine_similarity_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_ctc_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_dropout_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_elu_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_embedding_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_feature_alpha_dropout_with_train_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_feature_alpha_dropout_without_train_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_feature_alpha_dropout_without_train_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_feature_alpha_dropout_without_train_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_fractional_max_pool2d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_fractional_max_pool3d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_grid_sample_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_grid_sample_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_hardshrink_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_hardsigmoid_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_hardsigmoid_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_hardtanh_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_hinge_embedding_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_huber_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_huber_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_huber_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_instance_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_interpolate_bicubic_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_interpolate_linear_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_interpolate_linear_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_interpolate_nearest-exact_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_interpolate_trilinear_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_interpolate_trilinear_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_interpolate_trilinear_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_kl_div_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_l1_loss_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_layer_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_leaky_relu_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_local_response_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_local_response_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_logsigmoid_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_margin_ranking_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_margin_ranking_loss_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_max_pool2d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_max_pool3d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_max_unpool1d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_max_unpool1d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_max_unpool1d_grad_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_max_unpool2d_grad_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_mish_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_mish_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_mse_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_multilabel_soft_margin_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_nll_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_normalize_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pad_circular_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pad_circular_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pad_constant_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pad_reflect_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pad_replicate_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pad_replicate_negative_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pad_replicate_negative_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pad_replicate_negative_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pairwise_distance_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pixel_shuffle_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pixel_shuffle_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pixel_shuffle_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pixel_shuffle_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pixel_unshuffle_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pixel_unshuffle_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_pixel_unshuffle_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_poisson_nll_loss_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_poisson_nll_loss_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_relu6_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_relu6_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_relu6_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_relu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_relu_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_relu_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_rms_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_silu_complex_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_silu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_smooth_l1_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_soft_margin_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_soft_margin_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_softmin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_softmin_with_dtype_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_softmin_with_dtype_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_softplus_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_softshrink_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_softsign_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_softsign_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_tanhshrink_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_triplet_margin_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_triplet_margin_loss_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_triplet_margin_loss_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_triplet_margin_with_distance_loss_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_triplet_margin_with_distance_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_unfold_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nn_functional_unfold_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nonzero_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nonzero_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_nonzero_static_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_norm_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_norm_fro_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_norm_fro_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_norm_inf_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_norm_inf_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_normal_in_place_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_normal_in_place_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_normal_in_place_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ones_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ones_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ones_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ones_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ones_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ones_like_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ones_like_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ones_like_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_outer_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_outer_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_outer_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_outer_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_pca_lowrank_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_permute_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_permute_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_permute_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_permute_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_polar_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_polygamma_polygamma_n_0_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_polygamma_polygamma_n_1_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_polygamma_polygamma_n_2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_polygamma_polygamma_n_2_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_polygamma_polygamma_n_3_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_polygamma_polygamma_n_3_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_polygamma_polygamma_n_4_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_polygamma_polygamma_n_4_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_positive_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_positive_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_pow_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_pow_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_pow_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_pow_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_prod_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_prod_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_prod_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_put_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_put_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_rad2deg_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_rad2deg_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_rad2deg_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_rad2deg_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_rand_like_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_rand_like_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_randint_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_randint_like_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_randint_like_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_randn_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ravel_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ravel_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ravel_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_ravel_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_real_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_real_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_real_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_real_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_reciprocal_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_reciprocal_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_remainder_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_remainder_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_remainder_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_renorm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_repeat_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_repeat_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_repeat_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_repeat_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_repeat_interleave_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_repeat_interleave_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_reshape_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_reshape_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_reshape_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_reshape_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_resize__cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_resize__cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_resize__cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_resize_as__cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_resize_as__cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_resize_as__cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_resolve_conj_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_resolve_conj_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_resolve_conj_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_resolve_conj_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_resolve_conj_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_resolve_neg_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_rot90_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_rot90_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_rot90_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_round_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_round_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_round_decimals_0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_round_decimals_3_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_round_decimals_3_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_rsqrt_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_rsqrt_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_rsqrt_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scalar_tensor_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scalar_tensor_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scalar_tensor_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scatter_add_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scatter_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scatter_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scatter_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scatter_reduce_amax_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scatter_reduce_amin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scatter_reduce_amin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scatter_reduce_amin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scatter_reduce_mean_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scatter_reduce_prod_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scatter_reduce_sum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_scatter_reduce_sum_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_searchsorted_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_searchsorted_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_searchsorted_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_select_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_select_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_select_scatter_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_select_scatter_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sgn_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_short_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sigmoid_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sigmoid_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sigmoid_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sigmoid_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sigmoid_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_signal_windows_cosine_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_signal_windows_general_cosine_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_signbit_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sinc_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sinc_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sinh_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_slice_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_slice_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_slice_scatter_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_slice_scatter_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_softmax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_softmax_with_dtype_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_softmax_with_dtype_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_softmax_with_dtype_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_softmax_with_dtype_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_softmax_with_dtype_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sort_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sparse_mm_reduce_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sparse_sampled_addmm_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sparse_sampled_addmm_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sparse_sampled_addmm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_bessel_j0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_bessel_j1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_bessel_j1_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_bessel_y0_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_bessel_y0_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_chebyshev_polynomial_t_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_chebyshev_polynomial_t_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_chebyshev_polynomial_u_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_chebyshev_polynomial_u_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_chebyshev_polynomial_v_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_chebyshev_polynomial_w_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_entr_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_entr_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_erfcx_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_hermite_polynomial_h_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_hermite_polynomial_h_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_hermite_polynomial_h_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_i0e_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_i0e_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_i0e_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_i1_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_i1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_i1e_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_legendre_polynomial_p_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_log_ndtr_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_log_ndtr_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_log_ndtr_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_log_ndtr_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_modified_bessel_i0_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_modified_bessel_i0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_modified_bessel_i0_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_modified_bessel_i0_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_modified_bessel_i1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_modified_bessel_k0_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_modified_bessel_k0_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_ndtr_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_ndtr_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_ndtr_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_ndtr_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_ndtri_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_polygamma_special_polygamma_n_0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_polygamma_special_polygamma_n_0_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_polygamma_special_polygamma_n_0_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_polygamma_special_polygamma_n_0_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_scaled_modified_bessel_k1_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_scaled_modified_bessel_k1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_scaled_modified_bessel_k1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_shifted_chebyshev_polynomial_t_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_shifted_chebyshev_polynomial_t_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_shifted_chebyshev_polynomial_u_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_shifted_chebyshev_polynomial_u_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_shifted_chebyshev_polynomial_u_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_shifted_chebyshev_polynomial_v_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_shifted_chebyshev_polynomial_v_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_shifted_chebyshev_polynomial_w_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_spherical_bessel_j0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_xlog1py_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_zeta_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_special_zeta_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_list_args_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_list_args_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_list_args_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_with_sizes_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_with_sizes_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_with_sizes_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_with_sizes_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_with_sizes_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_with_sizes_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_with_sizes_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_with_sizes_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_with_sizes_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_split_with_sizes_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sqrt_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sqrt_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sqrt_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sqrt_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_square_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_square_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_square_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_squeeze_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_squeeze_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_squeeze_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_std_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_std_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_std_mean_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_std_mean_unbiased_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sub_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sum_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sum_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_sum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_svd_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_svd_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_t_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_t_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_t_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_t_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_take_along_dim_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_take_along_dim_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_take_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_take_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_take_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_take_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_take_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_take_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_tan_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_tan_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_tan_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_tensor_split_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_tensor_split_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_tensordot_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_tile_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_tile_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_tile_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_tile_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_tile_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_to_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_to_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_to_sparse_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_to_sparse_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_to_sparse_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_to_sparse_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_to_sparse_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_topk_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_topk_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_torch_ops_aten__efficient_attention_forward_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_torch_ops_aten__flash_attention_forward_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_torch_ops_aten__safe_softmax_default_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_torch_ops_aten__safe_softmax_default_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_trace_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_trace_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_trace_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_transpose_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_trapezoid_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_trapezoid_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_trapz_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_trapz_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_trapz_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_trapz_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_triangular_solve_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_tril_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_tril_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_tril_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_triu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_triu_indices_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_true_divide_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_true_divide_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_true_divide_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_trunc_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unbind_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unbind_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unbind_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unbind_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unflatten_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unfold_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unfold_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unfold_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unfold_copy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unfold_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unfold_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unfold_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unfold_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_uniform_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_uniform_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unique_consecutive_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unique_consecutive_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unique_consecutive_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unsafe_chunk_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unsafe_chunk_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unsafe_split_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unsqueeze_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unsqueeze_copy_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unsqueeze_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unsqueeze_copy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unsqueeze_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unsqueeze_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unsqueeze_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_unsqueeze_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_var_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_var_mean_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_var_mean_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_var_mean_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_var_mean_unbiased_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_vdot_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_vdot_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_view_as_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_view_as_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_view_as_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_view_as_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_view_as_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_view_as_real_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_view_copy_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_view_copy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_view_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_view_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_view_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_view_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_vsplit_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_vsplit_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_vsplit_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_vsplit_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_vsplit_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_vsplit_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_vstack_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_vstack_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_vstack_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_vstack_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_where_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_where_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_where_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_xlogy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_xlogy_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_xlogy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_zero__cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_inplace_zero__cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_zero__cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_zeros_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_inplace_zeros_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_inplace_zeros_like_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_inplace_zeros_like_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_inplace_zeros_like_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_inplace_zeros_like_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_H_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_H_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_H_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_H_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace___getitem___cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace___radd___cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace___radd___cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace___radd___cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace___radd___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace___rand___cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace___rand___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace___rdiv___cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace___rdiv___cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace___rdiv___cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace___rmatmul___cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace___rmul___cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace___rmul___cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace___rmul___cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace___rmul___cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace___ror___cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace___ror___cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace___ror___cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace___rpow___cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace___rsub___cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace___rsub___cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__batch_norm_with_update_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__batch_norm_with_update_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__chunk_cat_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__chunk_cat_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_abs_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_acos_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_acos_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_acos_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_add_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_addcdiv_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_addcdiv_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_addcdiv_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_addcmul_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_addcmul_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_addcmul_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_asin_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_asin_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_asin_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_atan_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_atan_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_ceil_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_ceil_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_ceil_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_ceil_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_clamp_max_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_clamp_min_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_clamp_min_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_clamp_min_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_clamp_min_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_clamp_min_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_clamp_min_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_cos_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_cos_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_cos_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_cosh_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_cosh_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_div_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_div_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_div_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_div_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_erf_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_erf_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_erfc_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_erfc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_erfc_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_exp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_exp_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_exp_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_expm1_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_expm1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_floor_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_floor_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_frac_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_frac_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_lerp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_lerp_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_lerp_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_lgamma_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_lgamma_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_log10_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_log1p_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_log2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_log2_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_log2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_log2_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_log2_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_log_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_log_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_max_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_maximum_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_maximum_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_maximum_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_maximum_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_minimum_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_neg_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_neg_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_neg_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_norm_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_pow_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_pow_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_pow_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_round_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_round_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_round_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_sign_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_sign_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_sin_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_sinh_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_sinh_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_sqrt_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_sqrt_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_sqrt_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_sqrt_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_sub_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_sub_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_sub_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_sub_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_tan_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_tan_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_tanh_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_tanh_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_tanh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_trunc_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_trunc_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__foreach_zero_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__native_batch_norm_legit_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace__softmax_backward_data_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__unsafe_masked_index_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace__unsafe_masked_index_put_accumulate_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace__unsafe_masked_index_put_accumulate_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__upsample_bilinear2d_aa_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace__upsample_bilinear2d_aa_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_abs_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_abs_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_abs_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_acosh_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_acosh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_acosh_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_acosh_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_add_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_add_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_addbmm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_addcdiv_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_addcmul_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_addcmul_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_addcmul_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_addmv_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_addmv_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_addmv_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_addr_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_addr_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_alias_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_alias_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_all_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_allclose_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_allclose_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_amax_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_amax_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_amin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_aminmax_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_aminmax_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_angle_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_angle_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_any_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_any_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_any_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_arange_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_arange_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_argmin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_argmin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_argmin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_argsort_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_argwhere_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_argwhere_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_argwhere_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_as_strided_copy_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_as_strided_copy_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_as_strided_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_as_strided_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_as_strided_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_as_strided_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_as_strided_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_as_strided_partial_views_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_as_strided_partial_views_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_as_strided_partial_views_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_as_strided_partial_views_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_as_strided_partial_views_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_asin_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_asinh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atan2_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atan2_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atan2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atan_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atan_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atan_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atan_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atan_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atanh_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atanh_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atleast_1d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atleast_1d_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atleast_1d_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atleast_1d_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atleast_2d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atleast_2d_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atleast_3d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atleast_3d_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atleast_3d_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atleast_3d_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_atleast_3d_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_bfloat16_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_bincount_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_bitwise_and_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_bitwise_and_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_bitwise_and_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_bitwise_left_shift_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_bitwise_or_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_bitwise_or_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_bitwise_xor_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_block_diag_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_block_diag_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_block_diag_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_block_diag_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_bmm_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_bool_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_bool_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_bool_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_broadcast_tensors_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_broadcast_to_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_broadcast_to_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_bucketize_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_byte_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_byte_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cartesian_prod_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cartesian_prod_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cartesian_prod_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cartesian_prod_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cartesian_prod_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cat_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cat_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cat_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cdist_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cdist_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cdouble_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cdouble_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cdouble_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ceil_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cfloat_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_char_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_char_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cholesky_inverse_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cholesky_solve_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_chunk_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_chunk_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_clamp_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_clamp_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_clamp_max_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_clamp_max_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_clamp_max_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_clamp_min_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_clamp_min_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_clamp_min_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_clone_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_clone_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_column_stack_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_combinations_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_combinations_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_combinations_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_conj_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_conj_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_conj_physical_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_conj_physical_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_constant_pad_nd_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_constant_pad_nd_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_constant_pad_nd_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_constant_pad_nd_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_constant_pad_nd_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_contiguous_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_contiguous_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_corrcoef_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_corrcoef_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_corrcoef_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cos_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cos_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cos_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cos_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cos_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cos_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cos_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cosh_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cosh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cosh_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_count_nonzero_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cov_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cov_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cov_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cross_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cross_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cummax_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cummax_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cummin_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cummin_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cumprod_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cumsum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cumsum_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cumulative_trapezoid_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_cumulative_trapezoid_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_deg2rad_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_deg2rad_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diag_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diag_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diag_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diag_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diag_embed_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diag_embed_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diag_embed_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diag_embed_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diagflat_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diagflat_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diagflat_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diagflat_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diagonal_copy_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diagonal_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diagonal_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diagonal_copy_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diagonal_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diagonal_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diagonal_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diagonal_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diagonal_scatter_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diagonal_scatter_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diff_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diff_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_diff_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_digamma_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_digamma_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_digamma_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_div_floor_rounding_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_div_no_rounding_mode_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_div_no_rounding_mode_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_div_no_rounding_mode_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_div_no_rounding_mode_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_div_trunc_rounding_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_div_trunc_rounding_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_dot_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_double_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_double_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_double_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_dsplit_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_dsplit_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_dsplit_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_dsplit_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_dsplit_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_dsplit_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_dsplit_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_dstack_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_dstack_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_dstack_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_dstack_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_einsum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_empty_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_empty_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_empty_like_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_empty_like_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_empty_like_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_empty_like_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_empty_like_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_empty_strided_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_empty_strided_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_empty_strided_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_eq_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_erf_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_erf_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_erf_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_erfc_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_erfc_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_erfinv_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_erfinv_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_erfinv_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_erfinv_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_erfinv_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_exp2_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_exp2_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_exp2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_expand_as_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_expand_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_expand_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_expand_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_expand_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_expm1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_expm1_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_expm1_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_exponential_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_eye_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_eye_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_eye_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_eye_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_eye_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fft2_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fft2_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fft2_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fft2_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fft_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fft_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fft_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fft_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fftn_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fftshift_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fftshift_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fftshift_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fftshift_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fftshift_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_fftshift_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_hfft2_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_hfft2_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_hfft2_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_hfft_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_hfftn_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_hfftn_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_ifft2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_ifft2_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_ifft_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_ifft_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_ifft_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_ifft_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_ifft_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_ifftn_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_ifftshift_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_ifftshift_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_ihfft_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_ihfft_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_irfft2_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_irfft_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_irfftn_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_irfftn_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_irfftn_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_rfft2_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_rfft2_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_rfft_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_rfftn_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fft_rfftn_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fill_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_flatten_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_flatten_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_flatten_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_flatten_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_flip_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_flip_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_flip_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_flip_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fliplr_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fliplr_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fliplr_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_flipud_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_flipud_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_flipud_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_flipud_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_float_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_float_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_float_power_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fmax_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fmax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fmax_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fmin_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fmin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fmod_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fmod_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_fmod_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_frexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_full_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_full_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_full_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_full_like_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_gather_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_gather_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_gather_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_gather_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_gcd_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_gcd_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ge_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ge_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ge_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_gradient_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_gradient_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_gradient_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_grid_sampler_2d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_grid_sampler_2d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_gt_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_gt_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_half_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_half_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_half_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_half_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_heaviside_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_heaviside_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_histc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_histc_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_hsplit_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_hsplit_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_hsplit_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_hsplit_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_hsplit_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_hsplit_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_hsplit_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_hstack_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_hypot_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_hypot_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_i0_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_igammac_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_imag_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_add_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_fill_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_fill_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_fill_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_put_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_put_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_reduce_amax_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_reduce_amin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_reduce_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_reduce_mean_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_reduce_mean_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_reduce_prod_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_reduce_prod_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_reduce_prod_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_reduce_prod_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_select_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_index_select_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_int_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_int_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_int_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_int_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_int_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isclose_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isclose_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isclose_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isclose_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isfinite_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isfinite_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isin_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isin_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isinf_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isinf_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isinf_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isnan_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isnan_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isneginf_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_isreal_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_item_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_item_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_item_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_jiterator_2inputs_2outputs_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_jiterator_2inputs_2outputs_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_jiterator_2inputs_2outputs_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_jiterator_2inputs_2outputs_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_jiterator_4inputs_with_extra_args_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_jiterator_4inputs_with_extra_args_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_jiterator_binary_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_jiterator_binary_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_jiterator_binary_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_jiterator_binary_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_jiterator_binary_return_by_ref_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_jiterator_binary_return_by_ref_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_jiterator_binary_return_by_ref_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_kthvalue_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_kthvalue_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lcm_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ldexp_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ldexp_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ldexp_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_le_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_le_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_le_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_le_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_le_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lerp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lgamma_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lgamma_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lgamma_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lgamma_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lgamma_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_cholesky_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_cholesky_ex_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_cond_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_cross_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_cross_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_cross_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_cross_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_cross_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_det_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_det_singular_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_diagonal_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_eig_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_eigh_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_eigvals_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_eigvalsh_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_eigvalsh_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_householder_product_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_inv_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_inv_ex_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_ldl_factor_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_ldl_factor_ex_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_ldl_factor_ex_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_ldl_solve_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_lstsq_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_lstsq_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_lu_factor_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_lu_factor_ex_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_matrix_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_norm_subgradients_at_zero_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_norm_subgradients_at_zero_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_qr_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_qr_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_slogdet_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_slogdet_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_solve_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_solve_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_solve_ex_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_solve_triangular_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_svd_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_tensorinv_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_vander_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_vander_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_vecdot_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_vecdot_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_vecdot_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_vecdot_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_vecdot_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_vector_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linalg_vector_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linspace_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linspace_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linspace_tensor_overload_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linspace_tensor_overload_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linspace_tensor_overload_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linspace_tensor_overload_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_linspace_tensor_overload_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log10_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log1p_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log1p_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log1p_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log2_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log2_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log2_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log2_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log_normal_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log_normal_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log_softmax_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log_softmax_with_dtype_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log_softmax_with_dtype_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_log_softmax_with_dtype_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logaddexp2_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logaddexp_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logaddexp_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logdet_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logdet_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logdet_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logical_and_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logical_and_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logical_and_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logical_not_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logical_or_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logical_or_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logical_xor_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logical_xor_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logical_xor_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logit_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logit_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logspace_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logspace_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logsumexp_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_logsumexp_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_long_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_long_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_long_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lt_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lt_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lt_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lu_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lu_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lu_solve_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lu_solve_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_lu_unpack_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mH_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mT_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mT_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mT_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mT_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mT_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mT_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mT_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_amax_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_amax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_amin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_argmin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_argmin_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_cumprod_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_cumsum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_fill_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_fill_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_fill_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_log_softmax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_log_softmax_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_logaddexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_logsumexp_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_logsumexp_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_logsumexp_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_logsumexp_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_mean_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_mean_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_mean_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_median_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_prod_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_prod_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_scatter_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_scatter_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_select_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_select_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_std_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_std_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_std_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_sum_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_sum_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_sum_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_var_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_masked_var_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_matmul_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_max_binary_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_max_binary_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_max_pool2d_with_indices_backward_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_maximum_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_median_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_median_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_median_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_median_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_meshgrid_list_of_tensors_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_meshgrid_list_of_tensors_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_meshgrid_list_of_tensors_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_meshgrid_list_of_tensors_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_meshgrid_list_of_tensors_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_meshgrid_list_of_tensors_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_meshgrid_variadic_tensors_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_meshgrid_variadic_tensors_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_meshgrid_variadic_tensors_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_meshgrid_variadic_tensors_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_min_binary_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_min_binary_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_min_binary_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_min_binary_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_min_reduction_no_dim_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_min_reduction_no_dim_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_min_reduction_no_dim_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_min_reduction_with_dim_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_min_reduction_with_dim_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_minimum_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_minimum_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_minimum_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mode_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mode_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mode_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mode_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_movedim_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_movedim_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_msort_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_msort_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mul_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mul_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mul_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_multinomial_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_multinomial_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_multinomial_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mvlgamma_mvlgamma_p_1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mvlgamma_mvlgamma_p_3_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mvlgamma_mvlgamma_p_3_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mvlgamma_mvlgamma_p_3_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mvlgamma_mvlgamma_p_3_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_mvlgamma_mvlgamma_p_5_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nan_to_num_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nan_to_num_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nan_to_num_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nan_to_num_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nanmean_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nanquantile_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nansum_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nansum_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_narrow_copy_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_narrow_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_narrow_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_native_batch_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_native_dropout_backward_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_native_dropout_backward_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_native_layer_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ne_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ne_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ne_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ne_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_neg_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_new_empty_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_new_empty_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_new_empty_strided_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_new_empty_strided_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_new_full_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_new_full_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_new_full_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_new_full_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_new_full_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_new_ones_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_new_ones_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_new_zeros_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nextafter_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_adaptive_avg_pool3d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_adaptive_max_pool1d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_adaptive_max_pool3d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_avg_pool1d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_avg_pool1d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_batch_norm_without_cudnn_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_bilinear_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_bilinear_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_binary_cross_entropy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_binary_cross_entropy_with_logits_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_celu_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_celu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_channel_shuffle_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_channel_shuffle_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_channel_shuffle_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_channel_shuffle_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_channel_shuffle_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_conv2d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_conv3d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_conv_transpose1d_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_conv_transpose1d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_conv_transpose2d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_conv_transpose2d_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_conv_transpose2d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_conv_transpose2d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_conv_transpose3d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_conv_transpose3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_cosine_embedding_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_cosine_embedding_loss_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_cosine_embedding_loss_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_cross_entropy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_dropout_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_dropout_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_embedding_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_embedding_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_feature_alpha_dropout_with_train_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_feature_alpha_dropout_without_train_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_feature_alpha_dropout_without_train_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_feature_alpha_dropout_without_train_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_feature_alpha_dropout_without_train_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_feature_alpha_dropout_without_train_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_fractional_max_pool3d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_gaussian_nll_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_gelu_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_glu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_glu_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_glu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_glu_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_grid_sample_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_hardshrink_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_hardsigmoid_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_hardswish_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_hardtanh_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_hardtanh_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_hinge_embedding_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_huber_loss_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_huber_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_interpolate_area_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_interpolate_bilinear_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_interpolate_nearest_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_interpolate_trilinear_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_kl_div_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_kl_div_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_l1_loss_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_layer_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_leaky_relu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_leaky_relu_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_linear_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_linear_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_local_response_norm_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_local_response_norm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_local_response_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_logsigmoid_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_margin_ranking_loss_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_max_pool2d_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_max_pool3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_max_pool3d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_max_unpool1d_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_max_unpool1d_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_max_unpool1d_grad_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_max_unpool2d_grad_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_max_unpool3d_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_max_unpool3d_grad_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_max_unpool3d_grad_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_multi_head_attention_forward_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_multi_margin_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_multilabel_soft_margin_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_normalize_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_normalize_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pad_circular_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pad_constant_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pad_constant_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pad_constant_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pad_constant_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pad_reflect_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pad_replicate_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pad_replicate_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pad_replicate_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pad_replicate_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pad_replicate_negative_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pad_replicate_negative_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pad_replicate_negative_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pairwise_distance_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pairwise_distance_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pdist_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pixel_shuffle_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pixel_unshuffle_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pixel_unshuffle_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pixel_unshuffle_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pixel_unshuffle_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pixel_unshuffle_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_pixel_unshuffle_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_poisson_nll_loss_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_poisson_nll_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_poisson_nll_loss_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_poisson_nll_loss_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_prelu_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_relu6_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_relu6_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_relu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_relu_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_rms_norm_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_rrelu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_rrelu_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_scaled_dot_product_attention_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_selu_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_selu_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_smooth_l1_loss_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_softmin_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_softmin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_softplus_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_softplus_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_tanhshrink_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_tanhshrink_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_tanhshrink_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_threshold_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_threshold_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_triplet_margin_loss_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_triplet_margin_loss_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_triplet_margin_with_distance_loss_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_triplet_margin_with_distance_loss_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nn_functional_upsample_bilinear_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nonzero_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nonzero_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nonzero_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nonzero_static_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_nonzero_static_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_norm_fro_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_norm_fro_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_norm_inf_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_norm_inf_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_normal_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_normal_number_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ones_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ones_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ones_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ones_like_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ones_like_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ormqr_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_outer_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_outer_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_outer_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_pca_lowrank_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_pca_lowrank_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_permute_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_permute_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_pinverse_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_pinverse_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_pinverse_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_polygamma_polygamma_n_0_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_polygamma_polygamma_n_1_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_polygamma_polygamma_n_1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_polygamma_polygamma_n_3_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_polygamma_polygamma_n_3_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_polygamma_polygamma_n_4_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_positive_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_positive_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_pow_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_pow_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_pow_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_prod_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_put_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_put_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_put_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_qr_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_qr_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_quantile_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_rad2deg_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_rad2deg_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_rand_like_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_randint_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_randint_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_randint_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_randint_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_randint_like_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_randn_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_randn_like_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ravel_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ravel_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_ravel_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_real_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_real_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_real_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_real_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_real_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_reciprocal_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_reciprocal_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_reciprocal_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_remainder_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_repeat_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_repeat_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_repeat_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_repeat_interleave_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_reshape_as_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_reshape_as_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_reshape_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_reshape_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_reshape_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_resize__cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_resize__cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_resize_as__cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_resize_as__cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_resize_as__cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_resolve_conj_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_resolve_conj_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_resolve_neg_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_resolve_neg_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_resolve_neg_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_roll_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_roll_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_roll_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_rot90_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_rot90_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_rot90_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_rot90_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_round_decimals_0_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_round_decimals_3_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_rsqrt_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_rsqrt_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_rsqrt_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_rsub_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_rsub_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_rsub_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_scalar_tensor_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_scalar_tensor_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_scatter_add_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_scatter_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_scatter_reduce_amax_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_scatter_reduce_amin_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_scatter_reduce_mean_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_scatter_reduce_mean_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_scatter_reduce_mean_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_scatter_reduce_sum_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_scatter_reduce_sum_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_scatter_reduce_sum_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_scatter_reduce_sum_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_select_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_select_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_select_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_select_scatter_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sgn_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sgn_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_short_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_short_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_short_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sigmoid_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sigmoid_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sign_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_signal_windows_exponential_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_signal_windows_exponential_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_signal_windows_gaussian_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_signal_windows_general_hamming_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_signal_windows_hamming_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_signal_windows_nuttall_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_signbit_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_signbit_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_signbit_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sin_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sin_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sin_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sinc_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sinc_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sinh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sinh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sinh_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_slice_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_slice_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_slice_scatter_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_slice_scatter_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_softmax_with_dtype_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_softmax_with_dtype_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sort_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sort_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sort_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sort_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sort_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sparse_mm_reduce_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sparse_sampled_addmm_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sparse_sampled_addmm_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_airy_ai_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_airy_ai_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_airy_ai_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_bessel_j0_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_bessel_j0_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_bessel_j0_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_bessel_j1_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_bessel_j1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_bessel_y0_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_bessel_y0_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_bessel_y1_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_chebyshev_polynomial_t_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_chebyshev_polynomial_t_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_chebyshev_polynomial_u_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_chebyshev_polynomial_u_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_chebyshev_polynomial_u_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_chebyshev_polynomial_v_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_chebyshev_polynomial_v_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_chebyshev_polynomial_w_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_chebyshev_polynomial_w_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_entr_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_entr_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_erfcx_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_erfcx_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_hermite_polynomial_he_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_hermite_polynomial_he_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_hermite_polynomial_he_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_hermite_polynomial_he_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_i0e_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_i0e_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_i0e_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_i1_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_i1e_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_i1e_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_laguerre_polynomial_l_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_legendre_polynomial_p_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_legendre_polynomial_p_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_log_ndtr_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_modified_bessel_i1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_modified_bessel_i1_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_modified_bessel_i1_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_modified_bessel_k0_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_modified_bessel_k1_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_modified_bessel_k1_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_modified_bessel_k1_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_modified_bessel_k1_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_modified_bessel_k1_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_ndtr_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_ndtr_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_ndtri_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_polygamma_special_polygamma_n_0_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_polygamma_special_polygamma_n_0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_polygamma_special_polygamma_n_0_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_polygamma_special_polygamma_n_0_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_scaled_modified_bessel_k0_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_scaled_modified_bessel_k1_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_shifted_chebyshev_polynomial_t_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_shifted_chebyshev_polynomial_t_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_shifted_chebyshev_polynomial_t_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_shifted_chebyshev_polynomial_u_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_shifted_chebyshev_polynomial_v_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_spherical_bessel_j0_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_xlog1py_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_xlog1py_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_special_zeta_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_split_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_split_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_split_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_split_list_args_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_split_list_args_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_split_list_args_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_split_list_args_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_split_list_args_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_split_with_sizes_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_split_with_sizes_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_split_with_sizes_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_split_with_sizes_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sqrt_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sqrt_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_square_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_square_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_square_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_squeeze_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_squeeze_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_squeeze_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_squeeze_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_squeeze_multiple_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_squeeze_multiple_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_squeeze_multiple_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_stack_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_stack_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_stack_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_stack_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_std_mean_unbiased_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_std_unbiased_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_stft_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_stft_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sub_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sub_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sum_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sum_to_size_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sum_to_size_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_sum_to_size_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_svd_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_svd_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_svd_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_t_copy_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_t_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_take_along_dim_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_take_along_dim_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_take_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_take_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tan_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tan_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tan_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tan_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tan_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tanh_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tanh_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tanh_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tensor_split_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tensor_split_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tile_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tile_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_to_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_to_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_to_sparse_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_to_sparse_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_to_sparse_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_to_sparse_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_topk_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_torch__scaled_mm_cuda_float8_e4m3fn, test/test_meta.py::TestMetaCUDA::test_meta_outplace_torch_ops_aten__safe_softmax_default_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_torch_ops_aten__safe_softmax_default_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_torch_ops_aten__safe_softmax_default_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_trace_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_trace_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_trace_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_transpose_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_transpose_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_trapezoid_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_trapz_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_trapz_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_triangular_solve_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tril_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tril_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_tril_indices_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_triu_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_triu_indices_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_triu_indices_cuda_int64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_true_divide_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unbind_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unflatten_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unflatten_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unflatten_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unflatten_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unfold_copy_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unfold_copy_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unfold_copy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unfold_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unfold_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_uniform_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unique_consecutive_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unique_consecutive_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unsafe_chunk_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unsafe_chunk_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unsafe_chunk_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unsafe_chunk_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unsafe_split_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unsafe_split_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unsafe_split_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unsqueeze_copy_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unsqueeze_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unsqueeze_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_unsqueeze_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_var_mean_unbiased_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_var_mean_unbiased_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_var_unbiased_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_vdot_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_view_as_complex_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_view_copy_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_view_cuda_bool, test/test_meta.py::TestMetaCUDA::test_meta_outplace_view_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_view_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_view_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_vsplit_cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_vsplit_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_vsplit_cuda_float32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_vsplit_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_vsplit_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_vsplit_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_vstack_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_vstack_cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_vstack_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_where_cuda_complex32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_where_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_where_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_xlogy_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_xlogy_cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_zero__cuda_bfloat16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_zero__cuda_float64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_zero__cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_zero__cuda_uint8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_zeros_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_zeros_cuda_int16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_zeros_cuda_int8, test/test_meta.py::TestMetaCUDA::test_meta_outplace_zeros_like_cuda_complex128, test/test_meta.py::TestMetaCUDA::test_meta_outplace_zeros_like_cuda_complex64, test/test_meta.py::TestMetaCUDA::test_meta_outplace_zeros_like_cuda_float16, test/test_meta.py::TestMetaCUDA::test_meta_outplace_zeros_like_cuda_int32, test/test_meta.py::TestMetaCUDA::test_meta_outplace_zeros_like_cuda_int64, test/test_meta.py::TestMetaCUDA::test_nan_to_num_cuda 2024-08-06T21:38:02.5104691Z 2024-08-06T21:38:05.5879742Z Running test_ops 9/11 ... [2024-08-06 21:38:05.587448] 2024-08-06T21:38:05.5884493Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_ops.py', '-m', 'not serial', '--shard-id=9', '--num-shards=11', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:38:05.588012] 2024-08-06T21:43:38.8622501Z 2024-08-06T21:43:38.8625706Z test_ops 2/11 was successful, full logs can be found in artifacts with path test/test-reports/test_ops_2.11_a0911fee89d45ee0_.log 2024-08-06T21:43:38.9837147Z Running 3026 items in this shard: test/test_ops.py::TestCommonCUDA::test_compare_cpu_H_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu___rmod___cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_cauchy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_diagonal_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_expand_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_eye_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_hstack_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_igamma_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_igammac_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_new_full_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_nextafter_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_nn_functional_nll_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_repeat_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_reshape_as_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_rot90_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_unsqueeze_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_var_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_view_as_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_xlogy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_atan2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_bernoulli_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_combinations_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_cummax_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_dist_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_div_floor_rounding_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_dsplit_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_empty_strided_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_full_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_geometric_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_linalg_eig_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_linalg_householder_product_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_logdet_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_masked_cumprod_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_masked_scatter_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_max_pool2d_with_indices_backward_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_max_reduction_no_dim_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_mm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_narrow_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_native_batch_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_new_full_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nextafter_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_conv2d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_feature_alpha_dropout_without_train_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_interpolate_bicubic_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_max_pool3d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_max_unpool2d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_multi_head_attention_forward_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_pad_constant_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_pad_reflect_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_softmin_with_dtype_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nonzero_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_norm_fro_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_norm_nuc_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_normal_number_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_quantile_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_resolve_neg_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_scatter_reduce_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_select_scatter_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_to_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_uniform_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_unsqueeze_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_var_unbiased_cuda_float32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_abs_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_as_strided_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_asin_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_char_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_dstack_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_hstack_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_index_copy_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_log_softmax_with_dtype_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_masked_fill_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_nn_functional_conv2d_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_nn_functional_conv_transpose3d_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_prod_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_rsqrt_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_sub_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_unsafe_chunk_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_view_as_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_view_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_dtypes_T_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes___getitem___cuda, test/test_ops.py::TestCommonCUDA::test_dtypes___rmul___cuda, test/test_ops.py::TestCommonCUDA::test_dtypes___ror___cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__batch_norm_with_update_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs__conversions_char_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_alias_copy_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_allclose_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_arange_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_as_strided_copy_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_atan2_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_bitwise_or_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_div_trunc_rounding_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_dot_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_equal_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_erf_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_expand_copy_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_fft_rfft2_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_isfinite_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_lgamma_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_log1p_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_log2_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_log_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_log_normal_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_meshgrid_variadic_tensors_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_nn_functional_hinge_embedding_loss_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_nn_functional_relu_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_nn_functional_softshrink_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_nn_functional_tanhshrink_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_nn_functional_triplet_margin_loss_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_remainder_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_sigmoid_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_special_i0e_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_special_log_softmax_with_dtype_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_special_multigammaln_mvlgamma_p_1_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_square_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_sub_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_triu_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_trunc_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_xlogy_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_addcmul_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_addmm_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_amax_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_amin_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_asin_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_bincount_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_cat_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_cdist_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_char_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_cholesky_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_count_nonzero_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_erfc_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_erfinv_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_fft_rfft2_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_flip_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_flipud_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_half_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_heaviside_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_hstack_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_index_reduce_amax_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_inner_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_isneginf_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_jiterator_binary_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_lcm_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_linalg_ldl_factor_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_linalg_norm_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_logical_not_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_masked_amin_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_masked_var_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_median_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_minimum_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_mm_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_mv_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_native_dropout_backward_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_neg_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_alpha_dropout_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_instance_norm_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_max_unpool2d_grad_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_mish_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_normalize_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_rrelu_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_norm_inf_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_ormqr_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_permute_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_resolve_neg_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_rot90_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_round_decimals_3_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_scatter_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_scatter_reduce_mean_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_select_scatter_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_softmax_with_dtype_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_special_laguerre_polynomial_l_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_special_modified_bessel_i1_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_special_ndtr_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_special_scaled_modified_bessel_k1_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_square_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_unfold_copy_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_uniform_cuda, test/test_ops.py::TestCommonCUDA::test_errors___rand___cuda, test/test_ops.py::TestCommonCUDA::test_errors___rdiv___cuda, test/test_ops.py::TestCommonCUDA::test_errors_aminmax_cuda, test/test_ops.py::TestCommonCUDA::test_errors_as_strided_scatter_cuda, test/test_ops.py::TestCommonCUDA::test_errors_complex_cuda, test/test_ops.py::TestCommonCUDA::test_errors_fft_fftn_cuda, test/test_ops.py::TestCommonCUDA::test_errors_fft_irfft2_cuda, test/test_ops.py::TestCommonCUDA::test_errors_fft_rfftn_cuda, test/test_ops.py::TestCommonCUDA::test_errors_fmax_cuda, test/test_ops.py::TestCommonCUDA::test_errors_gradient_cuda, test/test_ops.py::TestCommonCUDA::test_errors_kthvalue_cuda, test/test_ops.py::TestCommonCUDA::test_errors_median_cuda, test/test_ops.py::TestCommonCUDA::test_errors_nn_functional_l1_loss_cuda, test/test_ops.py::TestCommonCUDA::test_errors_nn_functional_multi_margin_loss_cuda, test/test_ops.py::TestCommonCUDA::test_errors_scatter_cuda, test/test_ops.py::TestCommonCUDA::test_errors_special_chebyshev_polynomial_t_cuda, test/test_ops.py::TestCommonCUDA::test_errors_special_shifted_chebyshev_polynomial_u_cuda, test/test_ops.py::TestCommonCUDA::test_errors_triu_cuda, test/test_ops.py::TestCommonCUDA::test_multiple_devices___rdiv___cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices__chunk_cat_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_allclose_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_amax_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_argwhere_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_as_strided_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_as_strided_scatter_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_block_diag_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_bmm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_cdist_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_ceil_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_cholesky_solve_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_combinations_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_corrcoef_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_cummin_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_cumulative_trapezoid_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_diagonal_copy_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_erfc_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_fft_fftshift_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_fft_ifft2_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_fft_ihfft2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_fft_rfftn_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_fmax_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_frexp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_gather_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_igammac_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_index_reduce_mean_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_jiterator_binary_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_jiterator_unary_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_linalg_lu_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_linalg_pinv_singular_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_linalg_qr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_linalg_vander_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_linalg_vecdot_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_long_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_masked_amin_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_masked_std_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_max_reduction_no_dim_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_meshgrid_list_of_tensors_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_min_reduction_with_dim_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_mode_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_mul_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_narrow_copy_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nextafter_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_adaptive_max_pool1d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_adaptive_max_pool3d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_alpha_dropout_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_celu_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_conv_transpose3d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_feature_alpha_dropout_with_train_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_hinge_embedding_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_margin_ranking_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_max_unpool2d_grad_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_max_unpool3d_grad_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_normalize_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_pad_circular_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_pad_replicate_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_prelu_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_relu_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_triplet_margin_loss_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_upsample_nearest_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_polygamma_polygamma_n_1_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_rad2deg_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_ravel_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_resize_as__cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_resolve_neg_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_scatter_reduce_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_scatter_reduce_prod_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_scatter_reduce_prod_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_signal_windows_gaussian_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_sin_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_slice_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_special_chebyshev_polynomial_u_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_special_i0e_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_special_modified_bessel_k0_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_special_polygamma_special_polygamma_n_0_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_sum_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_t_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_tanh_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_tile_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_to_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_to_sparse_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_torch_ops_aten__efficient_attention_forward_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_unique_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_unsafe_split_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_view_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_vsplit_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_vstack_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_zeros_like_cuda_int64, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values___radd___cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_addr_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_all_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_angle_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_diag_embed_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_digamma_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_fft_fft_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_fft_ifftshift_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_fft_irfftn_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_fmax_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_isclose_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_isfinite_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_jiterator_2inputs_2outputs_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_masked_fill_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_max_binary_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_nan_to_num_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_new_empty_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_permute_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_polygamma_polygamma_n_3_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_sigmoid_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_softmax_with_dtype_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_special_chebyshev_polynomial_t_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_special_hermite_polynomial_h_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_special_legendre_polynomial_p_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_squeeze_multiple_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_sum_to_size_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_t_copy_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_to_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_tril_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_true_divide_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_unique_consecutive_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_view_as_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_zeros_like_cuda_bool, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_T_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples__unsafe_masked_index_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples__unsafe_masked_index_put_accumulate_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_add_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_as_strided_copy_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_as_strided_scatter_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_asin_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_atan_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_atanh_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_atleast_3d_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_bfloat16_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_block_diag_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_bmm_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_bool_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_broadcast_tensors_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_chunk_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_chunk_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_clamp_max_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_combinations_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_complex_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_constant_pad_nd_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_cov_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_cov_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_cross_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_cummax_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_cumprod_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_cumulative_trapezoid_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_deg2rad_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_diag_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_div_floor_rounding_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_double_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_dstack_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_einsum_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_empty_strided_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_empty_strided_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_eq_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_exp2_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fft_fftn_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fft_hfft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fft_ifftshift_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fft_ihfft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fft_ihfftn_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fft_irfft2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fft_rfft_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fft_rfftn_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_flatten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_flatten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fliplr_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fliplr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_floor_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fmin_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_full_like_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_histc_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_index_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_index_select_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_int_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_isfinite_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_jiterator_4inputs_with_extra_args_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_jiterator_binary_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_det_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_ldl_solve_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_lu_factor_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_multi_dot_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_pinv_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_tensorinv_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_tensorinv_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_tensorsolve_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_vander_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linspace_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linspace_tensor_overload_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_log10_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_logaddexp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_logdet_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_logical_and_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_logical_not_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_lu_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_argmin_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_cumprod_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_cumprod_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_cumsum_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_logsumexp_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_mean_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_scatter_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_std_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_var_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_var_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_max_binary_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_max_reduction_with_dim_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_meshgrid_variadic_tensors_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_min_reduction_with_dim_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_msort_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nanmean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nansum_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_neg_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_neg_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_new_empty_strided_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_adaptive_max_pool1d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_avg_pool1d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_cosine_similarity_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_ctc_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_embedding_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_fractional_max_pool3d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_grid_sample_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_interpolate_nearest_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_linear_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_max_unpool2d_grad_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_multilabel_soft_margin_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_pad_constant_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_pad_replicate_negative_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_pairwise_distance_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_relu_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_threshold_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_triplet_margin_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_triplet_margin_loss_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_triplet_margin_with_distance_loss_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_upsample_nearest_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_ormqr_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_polygamma_polygamma_n_4_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_randn_like_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_ravel_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_remainder_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_repeat_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_roll_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_round_decimals_0_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_round_decimals_neg_3_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_rsub_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_scatter_reduce_amin_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_scatter_reduce_mean_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_select_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_signal_windows_cosine_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_signal_windows_general_hamming_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_sparse_sampled_addmm_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_airy_ai_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_bessel_y0_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_chebyshev_polynomial_t_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_entr_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_erfcx_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_i0e_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_i1_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_i1e_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_modified_bessel_i1_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_scaled_modified_bessel_k1_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_split_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_split_with_sizes_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_squeeze_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_std_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_std_mean_unbiased_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_std_unbiased_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_svd_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_unfold_copy_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_unfold_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_view_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_vstack_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_xlogy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_zero__cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_zeros_like_cuda_float32, test/test_ops.py::TestCommonCUDA::test_numpy_ref_argwhere_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_cat_cuda_int64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_clamp_cuda_int64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_diff_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_equal_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_jiterator_2inputs_2outputs_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_numpy_ref_linalg_cross_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_numpy_ref_meshgrid_variadic_tensors_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_meshgrid_variadic_tensors_cuda_int64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_nn_functional_gelu_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_nn_functional_smooth_l1_loss_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_roll_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_roll_cuda_int64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_signal_windows_blackman_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_signal_windows_general_cosine_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_signal_windows_general_hamming_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_squeeze_multiple_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_out_H_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out___getitem___cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__native_batch_norm_legit_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs__conversions_bool_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs__conversions_complex_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs__conversions_long_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_asin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_bitwise_left_shift_cuda_int64, test/test_ops.py::TestCommonCUDA::test_out__refs_bucketize_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_cat_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_constant_pad_nd_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_count_nonzero_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_div_no_rounding_mode_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_dstack_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_erf_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_fft_ihfft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_flipud_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_float_power_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_hsplit_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_istft_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out__refs_item_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_linalg_vecdot_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_log2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_logspace_tensor_overload_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_neg_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_nn_functional_hardshrink_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_nn_functional_poisson_nll_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_nn_functional_relu6_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_pow_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_repeat_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_round_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_select_scatter_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_sinh_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_special_entr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_special_log_softmax_with_dtype_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_special_multigammaln_mvlgamma_p_1_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_special_ndtr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_special_zeta_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_squeeze_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_sub_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_to_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_trace_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_unflatten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_vstack_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_add_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_alias_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_all_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_arange_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_argmin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_bernoulli_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_bitwise_right_shift_cuda_int64, test/test_ops.py::TestCommonCUDA::test_out_cauchy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_clamp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_combinations_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_dstack_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_fft_fftshift_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_fft_hfft2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_fft_ifftshift_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_fft_irfftn_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_flip_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_fmin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_full_like_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_gather_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_grid_sampler_2d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_gt_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_half_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_igamma_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_index_reduce_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_isnan_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_isreal_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_jiterator_binary_return_by_ref_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_ldexp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_linalg_ldl_solve_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_linalg_lu_solve_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_linalg_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_linalg_norm_subgradients_at_zero_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_lt_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_masked_argmin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_masked_normalize_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_masked_var_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_max_reduction_with_dim_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_min_reduction_no_dim_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_min_reduction_with_dim_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_mm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_msort_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nanmedian_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_new_ones_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_adaptive_avg_pool1d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_conv2d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_cosine_embedding_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_elu_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_layer_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_max_unpool1d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_max_unpool3d_grad_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_mse_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_nll_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_silu_complex_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_smooth_l1_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_tanhshrink_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_pca_lowrank_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_put_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_acos_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_acosh_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_addcdiv_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_addcmul_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_addmv_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_alias_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_atan2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_cat_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_cummin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_erfinv_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_fft_hfft2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_fft_ifft2_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_fft_ihfft2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_fft_rfft2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_index_reduce_prod_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_linalg_cholesky_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_linalg_eig_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_linalg_lu_solve_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_linalg_matrix_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_linalg_pinv_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_linalg_pinv_hermitian_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_log_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_logspace_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_masked_select_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_mm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_mv_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_normal_number_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_quantile_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_round_decimals_0_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_scatter_reduce_amin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_scatter_reduce_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_scatter_reduce_sum_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_sgn_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_softmax_with_dtype_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_special_ndtr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_special_ndtri_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_special_xlog1py_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_square_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_sub_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_sub_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_take_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_unsqueeze_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_where_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_roll_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_searchsorted_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_sign_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_signal_windows_general_cosine_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_split_with_sizes_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_squeeze_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_svd_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_svd_lowrank_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_tanh_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_vdot_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_warning___rxor___cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_broadcast_to_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_diag_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_dot_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_expm1_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_fft_fftshift_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_fft_ifftshift_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_fill_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_fliplr_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_fmax_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_linalg_svd_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_linspace_tensor_overload_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_log_softmax_with_dtype_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_logspace_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_narrow_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_nn_functional_dropout_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_nn_functional_mish_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_pow_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_randn_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_reshape_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_rot90_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_special_entr_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_special_log_ndtr_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_special_logit_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_t_copy_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_to_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_trunc_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_var_mean_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_vstack_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_atleast_1d_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_bitwise_and_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_block_diag_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_bool_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_byte_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_cauchy_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_char_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_cholesky_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_dist_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_empty_strided_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_expand_as_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_expand_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_fft_fftn_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_fft_hfft_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_fft_hfftn_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_fft_ifftshift_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_fft_rfft_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_geometric_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_gradient_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_histc_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_igamma_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_index_reduce_mean_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_isreal_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_istft_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_lgamma_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_linalg_ldl_factor_ex_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_log_softmax_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_logaddexp2_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_logical_not_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_lu_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_masked_logaddexp_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_masked_softmax_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_masked_softmin_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_max_binary_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_meshgrid_variadic_tensors_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_min_binary_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nansum_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_narrow_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_adaptive_max_pool1d_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_celu_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_dropout3d_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_embedding_bag_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_embedding_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_max_pool1d_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_max_pool2d_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_max_pool3d_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_max_unpool3d_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_scaled_dot_product_attention_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_softplus_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_norm_fro_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_normal_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_prod_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_reshape_as_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_resolve_conj_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_select_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_signal_windows_hamming_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_signal_windows_hann_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_signbit_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_sinh_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_slice_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_sparse_sampled_addmm_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_special_modified_bessel_i0_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_sqrt_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_to_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_torch__scaled_mm_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_torch_ops_aten__safe_softmax_default_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_tril_indices_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_triu_indices_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_view_cuda, test/test_ops.py::TestCommonCUDA::test_out_zero__cuda_float32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_asin_cuda_int32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_asinh_cuda_bool, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_atanh_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_cos_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_div_no_rounding_mode_cuda_int16, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_div_no_rounding_mode_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_erfinv_cuda_int32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_erfinv_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_erfinv_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_expm1_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_i0_cuda_int32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_i0_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_i0_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_ldexp_cuda_int32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_lgamma_cuda_int32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_log1p_cuda_int32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_log1p_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_logit_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_masked_mean_cuda_int32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_masked_var_cuda_int8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_mvlgamma_mvlgamma_p_5_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_polygamma_polygamma_n_2_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_polygamma_polygamma_n_3_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_rad2deg_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_rad2deg_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_special_chebyshev_polynomial_t_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_special_chebyshev_polynomial_v_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_special_hermite_polynomial_he_cuda_int32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_special_legendre_polynomial_p_cuda_bool, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_special_shifted_chebyshev_polynomial_v_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_special_shifted_chebyshev_polynomial_w_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_sqrt_cuda_int16, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_tan_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_tanh_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_tanh_cuda_int8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_xlogy_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_T_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_T_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_bfloat16_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_bfloat16_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_bool_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_byte_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_byte_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_byte_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_byte_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_cdouble_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_cfloat_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_double_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_float_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_float_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_half_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_half_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_int_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_long_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_short_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_short_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_abs_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_abs_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_abs_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_acos_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_acos_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_acos_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_add_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_addcdiv_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_addcdiv_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_addcmul_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_addr_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_addr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_addr_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_addr_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_alias_copy_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_alias_copy_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_all_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_all_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_allclose_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_amax_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_amax_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_amax_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_amin_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_arange_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_arange_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_as_strided_partial_views_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_as_strided_partial_views_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_as_strided_partial_views_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_as_strided_scatter_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_asinh_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_asinh_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atan2_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atan_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atanh_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atanh_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atanh_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atleast_1d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atleast_2d_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atleast_2d_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atleast_3d_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atleast_3d_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_bitwise_and_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_bitwise_not_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_bitwise_or_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_bitwise_right_shift_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_broadcast_tensors_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_bucketize_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_cat_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_cat_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_ceil_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_chunk_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_clamp_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_clamp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_clamp_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_clamp_min_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_clone_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_column_stack_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_conj_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_constant_pad_nd_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_constant_pad_nd_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_contiguous_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_copysign_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_cos_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_count_nonzero_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_cumprod_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_cumsum_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_deg2rad_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_diag_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_diag_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_diagonal_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_diagonal_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_digamma_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_div_no_rounding_mode_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_dot_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_dsplit_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_dstack_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_dstack_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_dstack_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_empty_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_empty_like_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_empty_strided_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_eq_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_equal_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_erf_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_erfinv_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_erfinv_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_exp2_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_exp_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_exp_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_expand_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_expand_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_expand_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_expand_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_exponential_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_eye_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_fft_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_fftshift_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_hfft2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_hfft_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_hfft_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_hfft_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_ifft2_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_ifftn_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_ifftn_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_ifftshift_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_ifftshift_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_ihfft2_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_ihfft2_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_ihfft_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_irfftn_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_rfft2_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_rfft_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_rfftn_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fill_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fliplr_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fliplr_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_flipud_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_float_power_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_float_power_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_floor_divide_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fmax_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fmin_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_gcd_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_ge_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_ge_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_geometric_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_heaviside_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_hstack_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_hypot_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_i0_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_igamma_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_index_copy_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_index_copy_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_index_copy_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_index_fill_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isinf_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isinf_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isinf_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isinf_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isneginf_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isneginf_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isposinf_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isreal_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isreal_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isreal_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_lcm_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_le_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_le_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_lerp_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_linalg_cross_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_linalg_diagonal_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_linalg_diagonal_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_linalg_matrix_norm_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_linspace_tensor_overload_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_linspace_tensor_overload_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log10_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log1p_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log2_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log_normal_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log_normal_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log_softmax_with_dtype_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log_softmax_with_dtype_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logical_and_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logical_or_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logical_xor_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logical_xor_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logspace_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logsumexp_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_masked_fill_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_masked_fill_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_masked_fill_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_maximum_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_maximum_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_meshgrid_variadic_tensors_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_meshgrid_variadic_tensors_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_minimum_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_minimum_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_movedim_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_mul_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_narrow_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_narrow_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_ne_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_neg_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_new_empty_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_new_full_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_new_ones_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_new_ones_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_new_zeros_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_new_zeros_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_alpha_dropout_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_elu_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_hardtanh_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_layer_norm_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_margin_ranking_loss_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_mish_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_pixel_shuffle_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_pixel_shuffle_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_pixel_unshuffle_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_poisson_nll_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_poisson_nll_loss_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_relu_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_selu_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_softmin_with_dtype_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_softplus_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_softshrink_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_tanhshrink_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_normal_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_ones_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_permute_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_permute_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_pow_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_rad2deg_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_randn_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_ravel_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_ravel_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_real_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_reciprocal_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_remainder_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_repeat_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_repeat_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_reshape_as_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_roll_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_round_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_round_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_rsqrt_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_rsqrt_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_rsub_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sgn_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sgn_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sigmoid_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sigmoid_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sigmoid_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sign_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_signbit_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_signbit_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sin_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sinh_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_softmax_with_dtype_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_softmax_with_dtype_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_erfcx_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_erfcx_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_i1e_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_log_ndtr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_log_ndtr_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_log_softmax_with_dtype_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_multigammaln_mvlgamma_p_1_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_ndtr_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_ndtr_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_softmax_with_dtype_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_softmax_with_dtype_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_softmax_with_dtype_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_xlog1py_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_xlog1py_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_xlog1py_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sqrt_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_square_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_squeeze_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sub_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sum_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sum_to_size_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sum_to_size_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sum_to_size_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_t_copy_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_t_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_take_along_dim_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_take_along_dim_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_take_along_dim_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_tensor_split_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_trace_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_true_divide_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_true_divide_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_trunc_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_trunc_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unflatten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unflatten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unflatten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unfold_copy_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unfold_copy_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unsqueeze_copy_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unsqueeze_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unsqueeze_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unsqueeze_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_var_mean_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_view_as_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_view_as_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_view_copy_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_view_copy_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_view_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_vsplit_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_vsplit_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_vstack_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_where_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_xlogy_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_xlogy_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_xlogy_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_zeros_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_zeros_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_T_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_add_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_diagonal_copy_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_diagonal_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_div_trunc_rounding_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_fft_ifftn_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_fft_ihfftn_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_fft_irfft_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_ge_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_igammac_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_index_add_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_logaddexp_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_normal__in_place_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_bfloat16_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_bfloat16_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_bfloat16_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_bool_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_byte_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_cdouble_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_cdouble_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_cfloat_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_cfloat_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_chalf_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_double_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_double_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_float_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_float_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_int_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_int_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_long_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_acos_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_acosh_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_addcmul_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_addcmul_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_addr_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_all_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_amax_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_amax_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_amin_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_as_strided_copy_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_as_strided_copy_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_as_strided_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_as_strided_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_as_strided_partial_views_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_as_strided_partial_views_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_as_strided_partial_views_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_as_strided_partial_views_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_as_strided_scatter_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_as_strided_scatter_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_as_strided_scatter_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_asin_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_asinh_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_atan2_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_atan2_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_atan_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_atanh_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_atanh_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_atanh_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_atleast_1d_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_atleast_1d_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_atleast_2d_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_atleast_3d_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_bitwise_not_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_bitwise_xor_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_broadcast_shapes_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_broadcast_tensors_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_broadcast_tensors_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_broadcast_tensors_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_broadcast_to_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_bucketize_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_bucketize_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_cat_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_cauchy_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_ceil_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_chunk_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_clone_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_column_stack_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_column_stack_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_constant_pad_nd_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_constant_pad_nd_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_copysign_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_cos_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_cos_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_cosh_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_count_nonzero_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_cumprod_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_cumsum_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_cumsum_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_diag_embed_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_diag_embed_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_diag_embed_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_diagonal_copy_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_diagonal_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_diagonal_scatter_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_div_no_rounding_mode_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_div_trunc_rounding_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_dot_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_dsplit_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_dsplit_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_dstack_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_dstack_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_empty_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_empty_like_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_empty_strided_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_empty_strided_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_eq_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_eq_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_equal_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_erf_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_exp2_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_exp2_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_exp_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_expand_copy_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_eye_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_fft2_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_fft2_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_fft_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_fft_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_fft_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_fftn_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_ifft2_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_ifft_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_ifft_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_ifftn_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_ifftn_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_ihfft2_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_ihfft_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_ihfftn_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_irfft2_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_irfft2_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_irfft_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_irfft_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_irfftn_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_irfftn_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_rfftn_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_flatten_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_flatten_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_flip_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fliplr_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_float_power_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_float_power_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_float_power_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_floor_divide_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fmin_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fmod_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_gcd_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_ge_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_geometric_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_gt_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_hstack_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_hstack_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_index_add_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_index_add_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_index_select_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_isclose_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_isnan_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_isnan_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_item_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_lcm_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_lcm_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_le_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_le_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_linalg_cross_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_linalg_diagonal_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_linalg_diagonal_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_linalg_norm_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_linalg_vector_norm_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_linspace_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_log10_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_log_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_log_softmax_with_dtype_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_log_softmax_with_dtype_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logical_and_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logical_and_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logical_not_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logical_xor_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logspace_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logspace_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logspace_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logspace_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logspace_tensor_overload_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logsumexp_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logsumexp_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_masked_fill_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_masked_fill_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_mean_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_mean_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_meshgrid_list_of_tensors_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_meshgrid_variadic_tensors_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_meshgrid_variadic_tensors_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_meshgrid_variadic_tensors_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_minimum_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_minimum_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_movedim_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_mul_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_mul_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_narrow_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_narrow_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_narrow_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_narrow_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_ne_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_ne_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_new_empty_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_new_ones_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_new_ones_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_celu_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_channel_shuffle_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_channel_shuffle_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_gelu_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_glu_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_group_norm_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_mse_loss_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_pairwise_distance_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_pdist_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_pixel_shuffle_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_pixel_unshuffle_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_poisson_nll_loss_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_relu_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_softmax_with_dtype_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_softmin_with_dtype_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_softplus_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_tanhshrink_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_tanhshrink_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_threshold_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_norm_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_normal_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_normal_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_permute_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_positive_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_positive_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_positive_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_rad2deg_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_rad2deg_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_randn_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_randn_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_randn_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_real_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_real_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_reciprocal_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_remainder_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_reshape_as_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_reshape_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_rot90_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_round_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_rsqrt_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_rsub_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_rsub_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_select_scatter_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sgn_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sigmoid_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sign_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_signbit_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sin_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sinc_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_softmax_with_dtype_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_softmax_with_dtype_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_bessel_j1_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_bessel_j1_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_bessel_j1_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_entr_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_erfcx_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_i0e_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_i0e_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_i1e_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_log_softmax_with_dtype_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_log_softmax_with_dtype_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_log_softmax_with_dtype_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_log_softmax_with_dtype_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_multigammaln_mvlgamma_p_1_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_multigammaln_mvlgamma_p_1_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_multigammaln_mvlgamma_p_3_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_multigammaln_mvlgamma_p_3_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_multigammaln_mvlgamma_p_5_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_ndtri_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_softmax_with_dtype_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_softmax_with_dtype_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sqrt_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sqrt_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_squeeze_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_stack_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_stack_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_std_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_std_mean_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sub_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sub_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sum_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sum_to_size_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_t_copy_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_t_copy_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_t_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_take_along_dim_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_take_along_dim_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_tanh_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_tensor_split_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_trace_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_trace_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_transpose_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_transpose_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_tril_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_tril_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_true_divide_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_true_divide_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_trunc_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unbind_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unbind_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unbind_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unflatten_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unfold_copy_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unfold_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unsqueeze_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unsqueeze_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_view_as_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_view_copy_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_view_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_vstack_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_where_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_where_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_where_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_xlogy_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_T_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_T_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_bool_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_byte_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_byte_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_cdouble_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_cdouble_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_cfloat_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_cfloat_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_double_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_float_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_long_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_short_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_short_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_short_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_abs_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_abs_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_abs_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_abs_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_acos_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_acos_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_add_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_add_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_addcdiv_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_addr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_alias_copy_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_alias_copy_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_all_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_all_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_amax_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_amax_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_any_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_arange_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_arange_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_as_strided_copy_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_as_strided_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_as_strided_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_as_strided_partial_views_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_as_strided_partial_views_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_asin_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atan2_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atan2_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atan2_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atan_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atan_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atan_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atanh_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atleast_2d_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atleast_2d_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atleast_2d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atleast_2d_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atleast_3d_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atleast_3d_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_bitwise_and_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_bitwise_or_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_bitwise_right_shift_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_block_diag_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_block_diag_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_broadcast_tensors_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_broadcast_tensors_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_broadcast_to_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_bucketize_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_cat_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_cat_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_cat_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_cauchy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_chunk_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_chunk_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_clamp_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_clamp_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_clamp_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_clamp_max_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_clone_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_column_stack_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_conj_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_conj_physical_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_constant_pad_nd_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_constant_pad_nd_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_contiguous_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_cos_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_cos_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_cos_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_cosh_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_count_nonzero_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_diag_embed_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_diagonal_copy_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_diagonal_scatter_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_digamma_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_div_floor_rounding_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_div_floor_rounding_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_div_trunc_rounding_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_div_trunc_rounding_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_div_trunc_rounding_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_dsplit_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_dstack_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_empty_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_empty_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_empty_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_empty_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_empty_like_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_empty_like_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_empty_strided_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_eq_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_eq_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_equal_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_erfc_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_exp2_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_expand_as_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_expand_as_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_expand_copy_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_expand_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_expand_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_expm1_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_expm1_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_expm1_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_exponential_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_exponential_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_fft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_fft_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_fftn_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_fftn_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_fftshift_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_hfft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_hfft_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_hfftn_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_hfftn_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_ifft2_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_ifft_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_ifftshift_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_ihfft2_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_ihfft2_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_ihfft_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_ihfft_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_ihfft_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_ihfftn_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_irfft2_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_irfft_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_rfft2_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_rfft_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_rfft_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fill_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fill_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_flatten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_flip_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_flip_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_flip_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fliplr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_flipud_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_flipud_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_flipud_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_float_power_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_floor_divide_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fmin_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fmod_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_ge_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_hsplit_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_hsplit_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_hypot_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_igammac_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_index_copy_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_isclose_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_isclose_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_isclose_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_isfinite_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_isfinite_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_isneginf_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_isneginf_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_le_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_linalg_matrix_norm_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_linalg_svd_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_linalg_vecdot_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_log10_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_log1p_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_log_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_log_normal_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logaddexp_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logical_and_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logical_not_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logical_or_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logical_xor_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logspace_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logspace_tensor_overload_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logspace_tensor_overload_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_masked_fill_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_masked_fill_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_meshgrid_list_of_tensors_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_minimum_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_movedim_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_mul_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_mul_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nan_to_num_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_narrow_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_narrow_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_ne_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_new_empty_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_new_ones_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_new_ones_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_new_zeros_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_new_zeros_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_new_zeros_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_dropout_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_glu_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_hardshrink_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_hardtanh_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_hinge_embedding_loss_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_l1_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_log_softmax_with_dtype_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_log_softmax_with_dtype_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_log_softmax_with_dtype_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_margin_ranking_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_mish_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_relu6_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_relu6_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_relu_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_softmax_with_dtype_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_softmin_with_dtype_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_tanhshrink_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_tanhshrink_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_threshold_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_triplet_margin_loss_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_normal__in_place_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_normal_number_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_permute_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_positive_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_prod_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_randn_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_ravel_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_real_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_reciprocal_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_reshape_as_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_roll_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_roll_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_rot90_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_rot90_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_round_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_rsqrt_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_rsub_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_select_scatter_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_select_scatter_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sgn_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sgn_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sign_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_signbit_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sin_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sinc_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sinh_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sinh_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sinh_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_softmax_with_dtype_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_entr_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_entr_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_erfcx_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_i0e_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_i1_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_i1e_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_logit_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_multigammaln_mvlgamma_p_1_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_multigammaln_mvlgamma_p_3_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_ndtri_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_ndtri_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_ndtri_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_softmax_with_dtype_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_xlog1py_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_zeta_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sqrt_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sqrt_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_squeeze_multiple_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_squeeze_multiple_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_stack_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_std_mean_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sub_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sum_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sum_to_size_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_t_copy_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_take_along_dim_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_take_along_dim_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_tanh_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_tanh_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_to_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_to_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_trace_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_transpose_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_tril_indices_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_true_divide_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_unfold_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_unfold_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_unfold_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_unfold_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_unsqueeze_copy_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_unsqueeze_copy_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_unsqueeze_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_unsqueeze_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_var_mean_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_view_as_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_view_copy_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_where_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_zeros_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_bfloat16_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_bfloat16_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_bool_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_byte_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_cfloat_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_cfloat_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_char_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_char_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_complex_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_double_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_double_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_float_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_half_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_half_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_int_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_long_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_long_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_long_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_abs_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_acos_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_acosh_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_acosh_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_add_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_addcmul_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_addcmul_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_addr_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_addr_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_alias_copy_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_alias_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_all_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_amax_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_amin_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_amin_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_any_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_arange_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_arange_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_arange_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_as_strided_copy_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_as_strided_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_asinh_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_asinh_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_atan2_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_atan_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_atan_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_atleast_1d_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_atleast_1d_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_atleast_3d_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_atleast_3d_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_bitwise_and_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_bitwise_left_shift_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_bitwise_not_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_block_diag_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_broadcast_tensors_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_broadcast_tensors_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_cat_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_chunk_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_clamp_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_clamp_max_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_clamp_min_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_clone_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_clone_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_conj_physical_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_constant_pad_nd_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_constant_pad_nd_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_constant_pad_nd_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_contiguous_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_contiguous_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_contiguous_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_cumprod_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_cumprod_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_cumprod_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_diag_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_diag_embed_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_diagonal_scatter_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_diagonal_scatter_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_div_floor_rounding_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_div_no_rounding_mode_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_div_trunc_rounding_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_dot_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_dstack_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_dstack_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_dstack_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_empty_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_empty_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_empty_like_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_empty_strided_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_equal_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_equal_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_equal_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_erf_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_erfc_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_exp2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_exp2_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_exp_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_exp_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_expand_as_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_expand_as_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_expand_copy_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_expand_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_expand_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_fft_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_fftshift_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_hfft2_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_hfftn_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_ifft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_ifftn_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_ifftshift_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_ifftshift_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_ihfft2_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_ihfft2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_irfft2_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_irfft_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_rfft2_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_rfft_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_rfft_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fill_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fliplr_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_float_power_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_float_power_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fmax_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fmin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fmin_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fmin_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fmod_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fmod_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fmod_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_frac_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_ge_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_geometric_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_geometric_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_gt_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_gt_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_heaviside_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_heaviside_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_heaviside_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_heaviside_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_heaviside_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_heaviside_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_heaviside_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_hsplit_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_hsplit_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_hsplit_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_hsplit_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_hstack_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_hstack_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_hstack_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_i0_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_imag_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_index_add_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_index_copy_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_index_fill_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_index_select_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_isinf_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_isinf_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_isinf_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_isnan_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_item_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_lgamma_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_linalg_diagonal_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_linalg_diagonal_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_linalg_norm_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_linalg_svd_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_linspace_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_linspace_tensor_overload_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_log10_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_log1p_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_log1p_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_log2_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_log_softmax_with_dtype_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_log_softmax_with_dtype_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logaddexp2_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logaddexp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logical_and_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logical_and_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logical_not_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logspace_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logspace_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logspace_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logsumexp_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_masked_fill_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_maximum_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_meshgrid_variadic_tensors_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_movedim_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_narrow_copy_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_narrow_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_ne_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_neg_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_neg_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_empty_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_empty_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_empty_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_empty_strided_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_empty_strided_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_zeros_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_zeros_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_alpha_dropout_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_dropout_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_gelu_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_glu_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_glu_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_hardshrink_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_hardtanh_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_hardtanh_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_l1_loss_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_l1_loss_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_l1_loss_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_layer_norm_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_layer_norm_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_layer_norm_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_log_softmax_with_dtype_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_log_softmax_with_dtype_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_margin_ranking_loss_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_margin_ranking_loss_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_mish_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_pixel_shuffle_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_pixel_shuffle_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_pixel_unshuffle_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_pixel_unshuffle_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_poisson_nll_loss_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_relu6_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_smooth_l1_loss_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_softmin_with_dtype_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_threshold_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_threshold_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_normal_number_mean_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_permute_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_positive_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_positive_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_positive_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_rad2deg_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_randn_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_ravel_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_ravel_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_real_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_reciprocal_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_remainder_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_remainder_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_repeat_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_repeat_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_reshape_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_rot90_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_round_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_rsqrt_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_select_scatter_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sgn_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sgn_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sigmoid_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sigmoid_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sigmoid_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_signbit_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_signbit_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sin_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sin_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sinc_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_bessel_j0_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_bessel_j1_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_entr_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_erfcx_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_i0e_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_i1e_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_logit_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_multigammaln_mvlgamma_p_1_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_multigammaln_mvlgamma_p_3_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_softmax_with_dtype_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_xlog1py_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_zeta_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_zeta_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_zeta_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sqrt_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_square_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_square_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_squeeze_multiple_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_std_mean_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sub_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sub_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sub_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sub_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sum_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sum_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_t_copy_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_t_copy_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_t_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_t_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_t_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_take_along_dim_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_take_along_dim_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_take_along_dim_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_tan_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_tan_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_tan_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_tanh_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_tanh_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_tanh_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_to_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_trace_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_trace_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_transpose_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_tril_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_tril_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_tril_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_triu_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_trunc_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_trunc_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_unflatten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_unfold_copy_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_unfold_copy_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_unfold_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_unfold_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_unsqueeze_copy_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_unsqueeze_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_unsqueeze_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_vdot_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_vdot_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_view_copy_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_view_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_vsplit_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_vsplit_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_vsplit_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_vsplit_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_vstack_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_where_cuda_float64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_H_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager___getitem___cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager___radd___cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager___rsub___cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager__batch_norm_with_update_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_addcdiv_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_addcmul_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_block_diag_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_bool_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_broadcast_to_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_cosh_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_cross_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_diag_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_diagonal_scatter_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_diff_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_dist_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_double_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_exp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_fft_fft2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_fft_fftshift_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_fft_hfftn_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_fft_irfft2_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_flatten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_float_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_frac_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_hsplit_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_igamma_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_imag_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_index_copy_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_index_put_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_jiterator_unary_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_ldexp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_diagonal_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_eig_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_householder_product_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_inv_ex_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_ldl_factor_ex_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_lu_solve_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_matrix_rank_hermitian_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_solve_ex_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_solve_ex_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_svd_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_tensorinv_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_logaddexp2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_logdet_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_logical_xor_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_logical_xor_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_lu_unpack_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_mT_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_masked_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_masked_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_masked_normalize_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_masked_scatter_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_masked_select_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_masked_std_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_masked_std_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_maximum_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_movedim_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_native_batch_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_native_dropout_backward_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_new_empty_strided_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_adaptive_max_pool1d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_conv1d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_conv2d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_conv3d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_interpolate_trilinear_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_layer_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_multi_head_attention_forward_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_pixel_unshuffle_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_rms_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_scaled_dot_product_attention_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nonzero_static_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_ormqr_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_ormqr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_outer_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_pow_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_qr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_rand_like_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_randn_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_resize_as__cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_rot90_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_rot90_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_scatter_reduce_prod_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_short_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_sinh_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_sparse_sampled_addmm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_special_entr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_split_with_sizes_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_sqrt_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_std_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_std_unbiased_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_tan_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_trapezoid_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_unsafe_chunk_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_var_mean_unbiased_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_view_as_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward__segment_reduce_offsets_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_acosh_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_angle_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_atleast_2d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_atleast_3d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_bernoulli_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_bmm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_broadcast_to_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_digamma_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_dist_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_div_floor_rounding_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_expand_as_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_fft_ifftn_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_index_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_index_reduce_prod_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_linalg_lu_factor_ex_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_linalg_norm_subgradients_at_zero_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_masked_cumsum_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_masked_fill_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_masked_mean_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_masked_median_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_masked_select_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_minimum_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_native_batch_norm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_celu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_conv2d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_glu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_layer_norm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_mse_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_pad_constant_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_rot90_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_round_decimals_0_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_scatter_reduce_prod_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_squeeze_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_torch_ops_aten__safe_softmax_default_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_trapz_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_T_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input__native_batch_norm_legit_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input__segment_reduce_lengths_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_acos_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_add_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_cdist_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_constant_pad_nd_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_dsplit_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_empty_strided_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_exp_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_expand_as_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_expand_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_fft_ifftshift_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_frac_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_i0_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_jiterator_unary_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_kron_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_kthvalue_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_linalg_matrix_rank_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_linalg_matrix_rank_hermitian_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_linalg_slogdet_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_linalg_svdvals_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_log_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_lu_unpack_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_masked_argmax_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_masked_scatter_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_masked_var_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nanmean_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_adaptive_max_pool3d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_avg_pool3d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_conv1d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_cosine_similarity_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_dropout_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_embedding_bag_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_group_norm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_max_pool3d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_max_unpool3d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_pad_reflect_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_silu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_softshrink_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_norm_nuc_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_permute_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_polar_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_randint_like_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_rsub_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_select_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_select_scatter_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_signal_windows_cosine_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_signbit_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_sparse_mm_reduce_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_special_bessel_y0_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_special_bessel_y1_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_special_erfcx_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_special_hermite_polynomial_he_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_special_ndtri_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_special_shifted_chebyshev_polynomial_t_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_square_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_svd_lowrank_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_tensordot_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_true_divide_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_var_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_view_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad___getitem___cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad___rsub___cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_addcdiv_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_amax_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_aminmax_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_atleast_2d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_cos_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_cross_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_cummin_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_diagonal_scatter_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_empty_like_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_eye_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_hsplit_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_index_put_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_linalg_det_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_linalg_lstsq_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_linalg_multi_dot_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_linalg_tensorsolve_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_linalg_vander_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_log10_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_log_softmax_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_masked_logsumexp_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_masked_var_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_max_reduction_no_dim_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_mm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_new_empty_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_new_empty_strided_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_new_ones_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_conv1d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_conv_transpose1d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_cosine_similarity_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_ctc_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_hardshrink_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_huber_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_kl_div_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_pad_circular_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_pad_constant_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_rrelu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_silu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_soft_margin_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_upsample_nearest_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_polygamma_polygamma_n_3_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_randint_like_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_resolve_conj_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_rsqrt_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_select_scatter_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_signal_windows_bartlett_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_sparse_mm_reduce_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_special_chebyshev_polynomial_w_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_stack_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_std_mean_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_topk_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_transpose_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_unflatten_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_H_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator___rmatmul___cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator___rmod___cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator__native_batch_norm_legit_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator__segment_reduce_offsets_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_addmv_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_argwhere_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_as_strided_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_as_strided_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_cdist_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_cummin_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_exp2_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_fft_fft_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_fft_ifft_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_index_fill_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_isclose_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linalg_det_singular_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linalg_eig_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linalg_householder_product_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linalg_inv_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linalg_matrix_power_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linalg_norm_subgradients_at_zero_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linalg_slogdet_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linalg_vector_norm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linspace_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_log_normal_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_logcumsumexp_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_mT_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_masked_amax_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_masked_argmin_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_masked_var_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_matrix_exp_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_mean_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_mvlgamma_mvlgamma_p_5_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_native_dropout_backward_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_new_zeros_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_conv_transpose3d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_cross_entropy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_hinge_embedding_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_interpolate_nearest-exact_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_max_unpool2d_grad_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_multilabel_margin_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_pairwise_distance_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_silu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_triplet_margin_with_distance_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_quantile_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_rand_like_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_real_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_reshape_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_roll_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_rsub_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_scatter_add_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_select_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_select_scatter_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_sign_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_signal_windows_general_hamming_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_slice_scatter_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_softmax_with_dtype_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_sparse_mm_reduce_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_special_entr_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_special_ndtr_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_special_scaled_modified_bessel_k0_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_special_scaled_modified_bessel_k1_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_special_shifted_chebyshev_polynomial_u_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_special_zeta_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_split_with_sizes_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_take_along_dim_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_tensor_split_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_tril_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_unique_consecutive_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_vstack_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_zero__cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_zeros_like_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay___radd___cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay__native_batch_norm_legit_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_aminmax_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_argmax_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_argsort_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_cdist_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_diagflat_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_div_floor_rounding_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_einsum_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_fft_ifft2_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_fft_ifft_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_fill_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_fliplr_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_flipud_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_hypot_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_index_reduce_mean_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_isneginf_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_le_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_linalg_cond_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_linalg_lstsq_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_linalg_lu_factor_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_linalg_pinv_hermitian_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_linalg_qr_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_linalg_tensorinv_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_log_softmax_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_logdet_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_logical_and_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_mT_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_masked_median_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_masked_scatter_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_masked_std_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_maximum_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_median_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_min_reduction_with_dim_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_mode_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_msort_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nan_to_num_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_narrow_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_adaptive_max_pool2d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_batch_norm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_conv3d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_conv_transpose3d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_dropout3d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_embedding_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_kl_div_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_logsigmoid_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_margin_ranking_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_max_pool2d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_max_pool3d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_max_unpool1d_grad_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_pairwise_distance_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_pixel_unshuffle_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_poisson_nll_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_softmin_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_softsign_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_tanhshrink_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_triplet_margin_with_distance_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_outer_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_polar_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_polygamma_polygamma_n_4_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_ravel_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_renorm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_resize_as__cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_roll_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_scatter_add_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_sign_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_slice_scatter_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_special_entr_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_special_shifted_chebyshev_polynomial_v_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_split_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_stack_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_std_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_std_mean_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_std_unbiased_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_stft_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_trapezoid_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_unsqueeze_copy_cuda_float32, test/test_ops.py::TestMathBitsCUDA::test_conj_view___radd___cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__chunk_cat_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs__conversions_bfloat16_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs__conversions_bool_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_acos_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_addr_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_broadcast_to_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_diag_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_empty_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_eq_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_eye_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_fft_ifft2_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_index_add_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_index_copy_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_item_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_linspace_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_log10_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_logical_and_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_new_empty_strided_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_nn_functional_pixel_shuffle_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_nn_functional_tanhshrink_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_squeeze_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_sub_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_true_divide_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_unbind_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_unsqueeze_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_vstack_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_as_strided_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_as_strided_scatter_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_broadcast_to_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_cholesky_solve_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_count_nonzero_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_empty_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_fft_hfftn_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_fill_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_index_copy_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_int_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_jiterator_binary_return_by_ref_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_lerp_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_linalg_norm_subgradients_at_zero_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_linalg_qr_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_logcumsumexp_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_logical_and_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_logical_or_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_masked_fill_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_masked_prod_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_masked_sum_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_matrix_exp_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_meshgrid_list_of_tensors_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_nn_functional_channel_shuffle_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_put_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_resolve_neg_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_select_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_sinc_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_square_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_take_along_dim_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_take_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_tile_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_unfold_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_vsplit_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs__conversions_byte_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs__conversions_cdouble_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs__conversions_chalf_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_abs_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_as_strided_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_asinh_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_cumsum_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_diagonal_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_empty_like_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_empty_strided_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_fft_ifft2_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_fft_ifftn_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_index_copy_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_isinf_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_linalg_vector_norm_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_narrow_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_nn_functional_log_softmax_with_dtype_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_nn_functional_softmin_with_dtype_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_randn_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_stack_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_stft_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_triu_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_atleast_3d_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_byte_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_combinations_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_conj_physical_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_cumulative_trapezoid_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_dist_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_dsplit_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_empty_permuted_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_eye_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_geqrf_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_hsplit_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_isreal_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_linalg_cond_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_linalg_det_singular_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_linalg_ldl_factor_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_logcumsumexp_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_masked_sum_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_nn_functional_pixel_unshuffle_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_nn_functional_silu_complex_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_norm_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_positive_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_pow_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_randn_like_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_repeat_interleave_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_rsqrt_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_select_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_short_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_sub_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_t_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_unsafe_chunk_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_where_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_view___rpow___cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs__conversions_chalf_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs__conversions_long_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_all_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_as_strided_copy_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_asin_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_block_diag_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_broadcast_tensors_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_ceil_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_conj_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_erfinv_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_fft_hfftn_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_fft_ihfft2_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_ge_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_hsplit_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_igammac_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_linspace_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_logical_and_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_ne_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_new_empty_strided_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_nn_functional_elu_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_nn_functional_glu_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_nn_functional_smooth_l1_loss_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_normal_number_mean_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_prod_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_round_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_sinh_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_tanh_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_tril_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_atan_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_cat_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_combinations_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_diag_embed_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_div_no_rounding_mode_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_div_trunc_rounding_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_equal_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_fft_rfft2_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_fft_rfft_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_flip_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_geqrf_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_gt_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_hypot_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_igamma_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_index_reduce_mean_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_isin_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_jiterator_binary_return_by_ref_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_linalg_eig_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_linalg_matrix_rank_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_linalg_pinv_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_linalg_vecdot_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_linspace_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_logaddexp2_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_logical_not_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_logical_xor_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nanmean_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_ne_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_new_ones_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_batch_norm_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_binary_cross_entropy_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_binary_cross_entropy_with_logits_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_conv1d_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_conv_transpose1d_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_grid_sample_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_interpolate_nearest_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_leaky_relu_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_max_unpool1d_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_max_unpool3d_grad_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_smooth_l1_loss_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_softsign_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_normal_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_normal_number_mean_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_polygamma_polygamma_n_2_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_prod_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_rand_like_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_resolve_conj_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_round_decimals_0_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_scatter_reduce_sum_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_sinc_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_special_bessel_j1_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_special_chebyshev_polynomial_t_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_special_i0e_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_special_i1e_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_special_shifted_chebyshev_polynomial_u_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_stack_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_std_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_stft_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_trapz_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_view_as_complex_cuda_float64, test/test_ops.py::TestFakeTensorCUDA::test_fake___radd___cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_addmm_decomposed_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_argsort_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_as_strided_copy_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_as_strided_scatter_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_atan2_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_atan_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast___radd___cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_acos_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_aminmax_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_as_strided_scatter_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_atan2_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_baddbmm_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_broadcast_shapes_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_cartesian_prod_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_cholesky_inverse_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_clamp_min_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_complex_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_copysign_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_corrcoef_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_div_trunc_rounding_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_exp_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_fft_hfft_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_grid_sampler_2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_jiterator_binary_return_by_ref_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_linalg_eigh_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_linalg_pinv_hermitian_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_linalg_solve_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_linalg_tensorsolve_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_logsumexp_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_masked_argmax_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_native_layer_norm_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_adaptive_avg_pool3d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_avg_pool2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_batch_norm_without_cudnn_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_conv_transpose2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_cosine_similarity_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_gaussian_nll_loss_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_interpolate_nearest_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_max_unpool1d_grad_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_max_unpool2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_poisson_nll_loss_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_relu6_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_relu_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_silu_complex_cuda_complex64, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_softmin_with_dtype_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_norm_nuc_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_real_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_scatter_reduce_sum_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_sigmoid_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_sign_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_sinh_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_sort_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_special_bessel_y1_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_special_chebyshev_polynomial_v_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_special_legendre_polynomial_p_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_special_ndtr_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_special_polygamma_special_polygamma_n_0_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_special_scaled_modified_bessel_k1_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_special_xlog1py_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_special_zeta_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_sub_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_tan_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_to_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_trunc_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_vsplit_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_bitwise_or_cuda_int64, test/test_ops.py::TestFakeTensorCUDA::test_fake_block_diag_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_chalf_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_copysign_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp__unsafe_masked_index_put_accumulate_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_amin_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_as_strided_scatter_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_cholesky_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_cov_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_diff_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_fft_fftn_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_fft_rfft_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_frexp_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_index_reduce_amax_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_linalg_cholesky_ex_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_linalg_diagonal_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_linalg_lu_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_linalg_svd_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_log_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_lu_unpack_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_masked_amax_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_masked_amin_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_masked_fill_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_matmul_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_msort_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_avg_pool1d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_avg_pool3d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_bilinear_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_channel_shuffle_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_logsigmoid_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_max_unpool2d_grad_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_pdist_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_pixel_unshuffle_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_relu6_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_softmin_with_dtype_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_polar_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_prod_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_round_decimals_neg_3_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_sigmoid_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_special_i0e_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_std_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_take_along_dim_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_tensordot_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_tile_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_unbind_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_unsafe_chunk_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_add_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_atan2_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_corrcoef_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_diag_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_diagonal_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_erf_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_fft_fftshift_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_i0_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_index_reduce_amax_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_index_reduce_prod_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_linalg_eigvalsh_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_linalg_lu_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_linalg_qr_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_linalg_tensorinv_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_log_softmax_with_dtype_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_masked_fill_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_masked_median_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_masked_select_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_max_reduction_with_dim_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_min_reduction_no_dim_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_mm_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_mv_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nanmedian_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_native_layer_norm_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_adaptive_avg_pool2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_adaptive_avg_pool3d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_adaptive_max_pool3d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_avg_pool1d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_avg_pool2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_conv2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_hinge_embedding_loss_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_l1_loss_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_margin_ranking_loss_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_max_unpool3d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_normalize_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_prelu_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_relu_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_permute_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_resolve_conj_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_resolve_neg_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_sinc_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_split_with_sizes_copy_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_squeeze_multiple_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_svd_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_trace_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_var_mean_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_vstack_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_div_no_rounding_mode_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_empty_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_empty_strided_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_erfc_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_exp_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_fft_ifftn_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_flip_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_fmax_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_index_reduce_mean_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_jiterator_unary_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_lcm_cuda_int64, test/test_ops.py::TestFakeTensorCUDA::test_fake_le_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_linalg_eig_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_linalg_ldl_solve_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_linalg_solve_ex_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_linalg_solve_triangular_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_log2_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_matmul_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_meshgrid_list_of_tensors_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_mm_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_mv_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nan_to_num_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_narrow_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_adaptive_avg_pool2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_alpha_dropout_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_channel_shuffle_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_conv3d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_cosine_embedding_loss_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_dropout2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_feature_alpha_dropout_with_train_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_max_unpool1d_grad_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_silu_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_normal_number_mean_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_quantile_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_randint_like_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_randn_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_scatter_add_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_scatter_reduce_mean_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_sigmoid_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_signal_windows_exponential_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_sin_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_softmax_with_dtype_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_sort_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_special_hermite_polynomial_h_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_special_modified_bessel_i1_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_special_modified_bessel_k1_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_special_polygamma_special_polygamma_n_0_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_triangular_solve_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_triu_indices_cuda_int64, test/test_ops.py::TestFakeTensorCUDA::test_fake_unfold_copy_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_where_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops__softmax_backward_data_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_abs_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_all_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_angle_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_as_strided_copy_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_baddbmm_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_bfloat16_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_bitwise_left_shift_cuda_int64, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_clamp_max_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_cummax_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_diag_embed_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_diff_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_fft_fft2_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_fft_ihfftn_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_flipud_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_frexp_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_full_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_geqrf_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_gradient_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_histc_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_igammac_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_imag_cuda_complex64, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_index_put_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_index_reduce_amin_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_index_reduce_prod_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_isnan_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_jiterator_4inputs_with_extra_args_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_linalg_cholesky_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_linalg_ldl_solve_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_linalg_vecdot_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_linspace_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_linspace_tensor_overload_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_logical_and_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_masked_logsumexp_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_masked_median_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_masked_norm_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_matmul_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_mode_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_narrow_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_native_dropout_backward_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_new_empty_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_new_zeros_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_nn_functional_multilabel_soft_margin_loss_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_norm_fro_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_ones_like_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_polygamma_polygamma_n_4_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_put_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_reshape_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_rot90_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_select_scatter_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_sign_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_signal_windows_blackman_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_softmax_with_dtype_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_squeeze_multiple_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_t_copy_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_tanh_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_tensor_split_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_trapz_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_unravel_index_cuda_int64, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_unsafe_chunk_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout__refs_arange_cuda_uint8, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout__refs_linspace_cuda_complex128, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout__refs_linspace_cuda_complex64, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout__refs_linspace_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout__refs_linspace_tensor_overload_cuda_int64, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout__refs_ones_cuda_int32, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout__refs_zeros_cuda_float16, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout__refs_zeros_cuda_uint8, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout_full_cuda_bfloat16, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout_logspace_cuda_complex64, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout_logspace_tensor_overload_cuda_int16, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout_logspace_tensor_overload_cuda_int32, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout_ones_cuda_bfloat16, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout_zeros_cuda_int16, test/test_ops.py::TestTagsCUDA::test_tags___rdiv___cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags___rxor___cuda_int64, test/test_ops.py::TestTagsCUDA::test_tags__refs__conversions_chalf_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_as_strided_copy_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_atan_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_broadcast_shapes_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_bucketize_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_diagonal_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_empty_strided_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_equal_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_expand_copy_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_expm1_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_fft_fft_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_fft_ihfftn_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_fill_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_floor_divide_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_gcd_cuda_int64, test/test_ops.py::TestTagsCUDA::test_tags__refs_i0_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_lcm_cuda_int64, test/test_ops.py::TestTagsCUDA::test_tags__refs_logical_and_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_masked_fill_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_nan_to_num_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_ne_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_nn_functional_elu_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_nn_functional_glu_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_nn_functional_l1_loss_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_nn_functional_layer_norm_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_nn_functional_prelu_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_nn_functional_relu6_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_nn_functional_threshold_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_ones_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_pow_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_reciprocal_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_softmax_with_dtype_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_special_ndtri_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_square_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_std_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_tril_indices_cuda_int64, test/test_ops.py::TestTagsCUDA::test_tags__refs_triu_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_vdot_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_view_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__segment_reduce_offsets_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_addcmul_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_addmv_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_argwhere_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_block_diag_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_cholesky_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_column_stack_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_contiguous_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_cross_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_diag_embed_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_dot_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_expand_copy_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_expand_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_fft_fftshift_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_fft_hfftn_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_flatten_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_grid_sampler_2d_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_index_reduce_amax_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_istft_cuda_complex64, test/test_ops.py::TestTagsCUDA::test_tags_jiterator_unary_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_lcm_cuda_int64, test/test_ops.py::TestTagsCUDA::test_tags_linalg_eig_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_linalg_pinv_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_logdet_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_logical_and_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_lu_solve_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_masked_amin_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_max_pool2d_with_indices_backward_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_max_reduction_with_dim_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nan_to_num_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nanmedian_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_narrow_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_native_dropout_backward_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_ne_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nextafter_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_batch_norm_without_cudnn_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_bilinear_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_binary_cross_entropy_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_feature_alpha_dropout_without_train_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_grid_sample_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_hardtanh_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_interpolate_nearest-exact_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_leaky_relu_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_max_unpool1d_grad_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_pad_replicate_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_prelu_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nonzero_static_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_norm_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_normal_number_mean_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_randint_like_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_resolve_conj_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_scatter_reduce_mean_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_scatter_reduce_sum_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_softmax_with_dtype_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_special_hermite_polynomial_h_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_special_polygamma_special_polygamma_n_0_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_special_zeta_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_split_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_split_with_sizes_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_std_mean_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_trapezoid_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_triangular_solve_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_unbind_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_unique_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_var_mean_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_view_cuda_float32 2024-08-06T21:43:39.0992443Z 2024-08-06T21:43:42.5232533Z Running test_module_tracker 1/1 ... [2024-08-06 21:43:42.522781] 2024-08-06T21:43:42.5237159Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_module_tracker.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:43:42.523266] 2024-08-06T21:43:46.6454331Z 2024-08-06T21:43:46.6456059Z test_module_tracker 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_module_tracker_1.1_1ed99d90ca62e7cc_.log 2024-08-06T21:43:46.6457849Z Running 2 items in this shard: test/test_module_tracker.py::TestModuleTracker::test_bw_detection, test/test_module_tracker.py::TestModuleTracker::test_module_hierarchy 2024-08-06T21:43:46.6458868Z 2024-08-06T21:43:50.2663535Z Running test_comparison_utils 1/1 ... [2024-08-06 21:43:50.265848] 2024-08-06T21:43:50.2666906Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_comparison_utils.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:43:50.266316] 2024-08-06T21:43:54.4391522Z 2024-08-06T21:43:54.4392778Z test_comparison_utils 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_comparison_utils_1.1_9801de73def40cda_.log 2024-08-06T21:43:54.4395438Z Running 5 items in this shard: test/test_comparison_utils.py::TestComparisonUtils::test_all_equal_no_assert, test/test_comparison_utils.py::TestComparisonUtils::test_all_equal_no_assert_nones, test/test_comparison_utils.py::TestComparisonUtils::test_assert_dtype, test/test_comparison_utils.py::TestComparisonUtils::test_assert_sizes, test/test_comparison_utils.py::TestComparisonUtils::test_assert_strides 2024-08-06T21:43:54.4397403Z 2024-08-06T21:43:58.0976167Z Running functorch/test_vmap 1/1 ... [2024-08-06 21:43:58.097160] 2024-08-06T21:43:58.0979863Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'functorch/test_vmap.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:43:58.097605] 2024-08-06T21:49:19.6482881Z 2024-08-06T21:49:19.6484304Z test_ops 9/11 was successful, full logs can be found in artifacts with path test/test-reports/test_ops_9.11_8c2f83fcb8d6dbc6_.log 2024-08-06T21:49:19.7992236Z Running 2966 items in this shard: test/test_ops.py::TestCommonCUDA::test_compare_cpu___rdiv___cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu___rmul___cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__native_batch_norm_legit_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs__conversions_chalf_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_as_strided_scatter_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_constant_pad_nd_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_fliplr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_hypot_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_linalg_svdvals_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_linspace_tensor_overload_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_log_normal_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_nn_functional_leaky_relu_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_normal_number_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_std_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_take_along_dim_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu__refs_unsqueeze_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_addcmul_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_addr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_cumsum_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_cumulative_trapezoid_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_diagonal_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_float_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_fmax_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_linalg_det_singular_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_logspace_tensor_overload_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_masked_logsumexp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_masked_select_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_median_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_dropout2d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_hardshrink_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_instance_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_interpolate_nearest-exact_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_pad_replicate_negative_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_nn_functional_relu6_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_qr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_rsub_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_slice_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_stft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_svd_lowrank_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_trace_cuda_float32, test/test_ops.py::TestCommonCUDA::test_compare_cpu_trapz_cuda_float32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_acosh_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_as_strided_partial_views_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_conj_physical_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_diag_embed_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_hsplit_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_isreal_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_nn_functional_conv_transpose1d_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_ones_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_rand_like_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_true_divide_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_unsafe_split_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_complex_half_reference_testing_unsqueeze_copy_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_dtypes___rdiv___cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_T_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs__conversions_cfloat_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_all_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_asinh_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_bitwise_xor_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_broadcast_tensors_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_cauchy_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_cos_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_deg2rad_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_diag_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_diagonal_copy_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_diagonal_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_empty_strided_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_erfc_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_fliplr_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_frexp_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_ne_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_nn_functional_huber_loss_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_nn_functional_leaky_relu_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_nn_functional_mse_loss_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_nn_functional_pairwise_distance_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_nn_functional_softmax_with_dtype_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_nn_functional_softmin_with_dtype_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_nn_functional_softplus_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_normal_number_mean_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_repeat_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_round_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_special_ndtri_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_trace_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_unfold_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_view_as_complex_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__refs_vstack_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes__segment_reduce_offsets_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_acos_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_bitwise_xor_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_byte_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_cartesian_prod_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_conj_physical_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_dot_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_dstack_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_fliplr_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_index_select_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_isclose_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_isinf_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_lerp_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_lgamma_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_linalg_cross_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_linalg_householder_product_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_linalg_inv_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_linalg_inv_ex_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_linalg_ldl_factor_ex_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_linalg_slogdet_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_logaddexp2_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_logit_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_mH_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_masked_amax_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_masked_argmax_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_masked_cumsum_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_masked_mean_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_masked_normalize_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_matrix_exp_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_mul_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nextafter_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_batch_norm_without_cudnn_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_ctc_loss_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_group_norm_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_margin_ranking_loss_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_max_unpool1d_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_max_unpool2d_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_pad_reflect_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_pad_replicate_negative_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nn_functional_pairwise_distance_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_nonzero_static_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_normal_in_place_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_polar_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_polygamma_polygamma_n_1_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_positive_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_rand_like_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_remainder_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_scatter_reduce_sum_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_signal_windows_blackman_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_signal_windows_nuttall_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_signbit_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_special_chebyshev_polynomial_t_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_special_hermite_polynomial_h_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_special_hermite_polynomial_he_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_special_shifted_chebyshev_polynomial_u_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_special_spherical_bessel_j0_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_special_zeta_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_stft_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_trunc_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_unique_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_unsafe_chunk_cuda, test/test_ops.py::TestCommonCUDA::test_dtypes_view_as_cuda, test/test_ops.py::TestCommonCUDA::test_errors_diagonal_copy_cuda, test/test_ops.py::TestCommonCUDA::test_errors_dstack_cuda, test/test_ops.py::TestCommonCUDA::test_errors_fft_hfft2_cuda, test/test_ops.py::TestCommonCUDA::test_errors_fft_rfft_cuda, test/test_ops.py::TestCommonCUDA::test_errors_gcd_cuda, test/test_ops.py::TestCommonCUDA::test_errors_hsplit_cuda, test/test_ops.py::TestCommonCUDA::test_errors_linspace_cuda, test/test_ops.py::TestCommonCUDA::test_errors_linspace_tensor_overload_cuda, test/test_ops.py::TestCommonCUDA::test_errors_log_normal_cuda, test/test_ops.py::TestCommonCUDA::test_errors_logical_xor_cuda, test/test_ops.py::TestCommonCUDA::test_errors_masked_fill_cuda, test/test_ops.py::TestCommonCUDA::test_errors_multinomial_cuda, test/test_ops.py::TestCommonCUDA::test_errors_narrow_copy_cuda, test/test_ops.py::TestCommonCUDA::test_errors_nn_functional_avg_pool3d_cuda, test/test_ops.py::TestCommonCUDA::test_errors_renorm_cuda, test/test_ops.py::TestCommonCUDA::test_errors_special_chebyshev_polynomial_u_cuda, test/test_ops.py::TestCommonCUDA::test_errors_special_shifted_chebyshev_polynomial_w_cuda, test/test_ops.py::TestCommonCUDA::test_errors_special_xlog1py_cuda, test/test_ops.py::TestCommonCUDA::test_errors_uniform_cuda, test/test_ops.py::TestCommonCUDA::test_errors_vdot_cuda, test/test_ops.py::TestCommonCUDA::test_errors_vstack_cuda, test/test_ops.py::TestCommonCUDA::test_errors_where_cuda, test/test_ops.py::TestCommonCUDA::test_multiple_devices_H_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices___rmul___cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices__unsafe_masked_index_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_addcdiv_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_addmm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_addr_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_any_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_argmax_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_bitwise_not_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_bitwise_or_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_block_diag_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_bucketize_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_cholesky_inverse_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_column_stack_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_cos_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_diag_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_diag_embed_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_diagonal_scatter_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_div_trunc_rounding_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_dstack_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_eq_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_erf_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_fft_fft2_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_fft_hfft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_fft_ifft_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_fliplr_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_half_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_heaviside_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_hstack_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_hstack_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_i0_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_isinf_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_isneginf_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_jiterator_2inputs_2outputs_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_jiterator_4inputs_with_extra_args_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_jiterator_binary_return_by_ref_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_ldexp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_linalg_multi_dot_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_linalg_tensorsolve_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_log2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_logdet_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_logspace_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_masked_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_masked_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_masked_softmin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_masked_sum_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_matmul_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_max_binary_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_movedim_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_mv_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_mvlgamma_mvlgamma_p_1_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_mvlgamma_mvlgamma_p_5_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_new_empty_strided_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_new_ones_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_gaussian_nll_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_hardtanh_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_interpolate_bilinear_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_max_unpool2d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_mse_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_pdist_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_pixel_shuffle_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_poisson_nll_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_relu_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_rms_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_selu_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_softmin_with_dtype_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_nn_functional_threshold_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_norm_fro_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_outer_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_prod_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_randint_like_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_repeat_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_repeat_interleave_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_scatter_reduce_amin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_signal_windows_general_cosine_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_slice_scatter_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_special_airy_ai_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_special_bessel_j0_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_special_bessel_y1_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_special_erfcx_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_special_hermite_polynomial_h_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_special_scaled_modified_bessel_k0_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_special_shifted_chebyshev_polynomial_w_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_special_zeta_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_split_with_sizes_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_sqrt_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_take_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_tril_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_unsafe_chunk_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_unsafe_split_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_unsqueeze_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_multiple_devices_view_as_cuda_int64, test/test_ops.py::TestCommonCUDA::test_multiple_devices_zero__cuda_float32, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_abs_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_atanh_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_char_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_chunk_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_clone_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_diagflat_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_empty_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_equal_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_fft_fft2_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_fft_hfftn_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_float_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_full_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_index_put_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_isneginf_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_jiterator_binary_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_log10_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_logsumexp_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_lt_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_meshgrid_list_of_tensors_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_min_reduction_no_dim_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_minimum_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_ne_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_nn_functional_pixel_shuffle_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_nn_functional_pixel_unshuffle_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_special_modified_bessel_k0_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_special_shifted_chebyshev_polynomial_t_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_special_shifted_chebyshev_polynomial_u_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_split_with_sizes_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_tan_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_torch_ops_aten__safe_softmax_default_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_unfold_cuda_bool, test/test_ops.py::TestCommonCUDA::test_non_standard_bool_values_view_cuda_bool, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_H_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples___rmatmul___cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples__chunk_cat_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples__unsafe_masked_index_put_accumulate_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_acosh_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_addr_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_asin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_atan_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_atanh_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_atleast_1d_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_baddbmm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_bfloat16_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_bitwise_xor_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_block_diag_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_broadcast_to_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_broadcast_to_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_cartesian_prod_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_ceil_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_cfloat_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_char_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_cosh_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_cov_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_cross_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_diagonal_scatter_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_diff_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_digamma_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_dist_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_div_no_rounding_mode_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_empty_like_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_empty_like_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_erfinv_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_expand_copy_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_expm1_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_eye_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fft_fftn_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fft_irfft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fft_rfft2_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fill_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_flatten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_fmod_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_ge_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_heaviside_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_hsplit_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_hstack_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_index_put_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_isclose_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_isinf_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_isreal_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_item_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_jiterator_2inputs_2outputs_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_kthvalue_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_kthvalue_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_ldexp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_cholesky_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_cholesky_ex_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_det_singular_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_eigvalsh_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_lstsq_grad_oriented_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_lu_factor_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_matrix_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_norm_subgradients_at_zero_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_linalg_vector_norm_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_log_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_logical_and_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_logit_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_logspace_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_logsumexp_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_amin_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_median_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_normalize_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_prod_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_masked_std_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_matrix_exp_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_msort_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_mul_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_mv_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_mvlgamma_mvlgamma_p_1_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nan_to_num_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_native_batch_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_ne_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_new_zeros_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nextafter_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_adaptive_avg_pool1d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_batch_norm_without_cudnn_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_binary_cross_entropy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_group_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_hardtanh_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_mish_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_multi_margin_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_normalize_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_pixel_unshuffle_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_silu_complex_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_softmin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_softmin_with_dtype_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_nn_functional_tanhshrink_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_ones_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_ones_like_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_permute_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_polygamma_polygamma_n_4_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_pow_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_real_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_resolve_neg_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_rsub_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_scatter_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_searchsorted_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_select_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_sign_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_signbit_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_sin_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_sinh_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_softmax_with_dtype_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_sparse_mm_reduce_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_bessel_j1_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_chebyshev_polynomial_u_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_chebyshev_polynomial_v_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_i1_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_laguerre_polynomial_l_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_special_xlog1py_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_split_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_split_list_args_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_sqrt_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_stack_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_sub_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_svd_lowrank_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_t_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_transpose_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_triangular_solve_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_unsafe_chunk_cuda_int64, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_view_as_cuda_float32, test/test_ops.py::TestCommonCUDA::test_noncontiguous_samples_vsplit_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_allclose_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_broadcast_tensors_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_clone_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_clone_cuda_int64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_linalg_vecdot_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_nn_functional_layer_norm_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_nn_functional_pairwise_distance_cuda_float64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_nn_functional_pairwise_distance_cuda_int64, test/test_ops.py::TestCommonCUDA::test_numpy_ref_nn_functional_rms_norm_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_numpy_ref_unbind_cuda_float64, test/test_ops.py::TestCommonCUDA::test_out__refs__conversions_bfloat16_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_add_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_allclose_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_diag_embed_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_fft_hfft2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_fft_ifft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_fill_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_flatten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_floor_divide_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_fmin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_frac_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_heaviside_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_isposinf_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_linalg_svdvals_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_logaddexp2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_minimum_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_new_ones_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_reshape_as_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_sigmoid_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_special_spherical_bessel_j0_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_special_xlog1py_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_squeeze_multiple_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_stft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_take_along_dim_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out__refs_tan_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_acosh_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_allclose_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_amax_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_amin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_as_strided_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_atleast_1d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_bincount_cuda_int64, test/test_ops.py::TestCommonCUDA::test_out_block_diag_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_bmm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_cholesky_inverse_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_corrcoef_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_diag_embed_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_div_trunc_rounding_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_empty_permuted_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_exponential_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_fft_fft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_fft_ihfftn_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_fmod_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_gcd_cuda_int64, test/test_ops.py::TestCommonCUDA::test_out_ge_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_isposinf_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_item_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_jiterator_4inputs_with_extra_args_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_le_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_lgamma_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_linalg_cross_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_linalg_det_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_linalg_eigvals_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_linalg_householder_product_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_linalg_ldl_factor_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_linalg_matrix_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_linalg_matrix_rank_hermitian_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_logical_and_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_masked_fill_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_max_binary_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_maximum_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_mvlgamma_mvlgamma_p_3_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_ne_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_adaptive_avg_pool3d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_batch_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_binary_cross_entropy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_conv_transpose2d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_instance_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_multi_margin_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_multilabel_soft_margin_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_pixel_shuffle_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nn_functional_poisson_nll_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_nonzero_static_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_normal_in_place_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_randn_like_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error__batch_norm_with_update_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_acos_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_acosh_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_addmm_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_addmv_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_amin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_angle_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_as_strided_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_asinh_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_asinh_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_baddbmm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_clamp_max_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_clamp_min_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_cumprod_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_diagonal_copy_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_dstack_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_fft_hfft2_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_fft_ifftn_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_fft_irfft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_fft_irfftn_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_float_power_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_float_power_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_i0_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_index_add_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_kron_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_linalg_cross_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_linalg_eigh_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_linalg_inv_ex_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_linalg_lstsq_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_linalg_solve_ex_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_linalg_vecdot_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_linspace_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_logcumsumexp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_lu_solve_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_max_reduction_no_dim_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_minimum_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_mvlgamma_mvlgamma_p_3_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_nanmean_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_nn_functional_linear_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_nn_functional_logsigmoid_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_nn_functional_normalize_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_norm_inf_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_pow_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_rad2deg_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_remainder_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_renorm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_scatter_add_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_scatter_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_scatter_reduce_prod_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_sqrt_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_topk_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_trunc_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_requires_grad_error_view_copy_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_out_rot90_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_special_legendre_polynomial_p_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_special_ndtr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_special_scaled_modified_bessel_k0_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_split_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_tril_indices_cuda_int64, test/test_ops.py::TestCommonCUDA::test_out_unbind_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_var_mean_unbiased_cuda_float32, test/test_ops.py::TestCommonCUDA::test_out_warning___getitem___cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs__conversions_cdouble_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs__conversions_half_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_alias_copy_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_ceil_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_digamma_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_div_trunc_rounding_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_flipud_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_frac_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_hsplit_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_lcm_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_linalg_vector_norm_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_logaddexp_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_logical_and_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_maximum_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_nn_functional_l1_loss_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_nn_functional_softmin_with_dtype_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_nn_functional_softplus_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_nn_functional_softshrink_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_ones_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_reshape_as_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_sgn_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_sigmoid_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_sign_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_special_log_softmax_with_dtype_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_std_mean_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_tensor_split_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_transpose_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_unbind_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_unsqueeze_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning__refs_zeros_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_cartesian_prod_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_cat_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_clamp_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_combinations_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_copysign_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_corrcoef_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_cumprod_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_double_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_dsplit_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_dstack_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_empty_permuted_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_eq_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_erf_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_expand_copy_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_fill_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_flipud_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_float_power_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_frac_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_index_reduce_prod_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_jiterator_binary_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_le_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_linalg_diagonal_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_linalg_qr_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_linalg_slogdet_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_logspace_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_masked_argmax_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_masked_argmin_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_matrix_exp_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_maximum_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_mvlgamma_mvlgamma_p_3_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nanquantile_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_narrow_copy_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_native_dropout_backward_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_ne_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_new_ones_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_huber_loss_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_layer_norm_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_max_unpool2d_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_pad_reflect_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_nn_functional_pairwise_distance_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_norm_nuc_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_pinverse_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_polygamma_polygamma_n_1_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_quantile_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_randn_like_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_scatter_reduce_sum_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_signal_windows_nuttall_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_sort_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_sparse_mm_reduce_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_special_bessel_j0_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_special_chebyshev_polynomial_t_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_special_chebyshev_polynomial_v_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_special_shifted_chebyshev_polynomial_v_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_special_xlog1py_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_squeeze_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_trapz_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_unsafe_split_cuda, test/test_ops.py::TestCommonCUDA::test_out_warning_where_cuda, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float___rdiv___cuda_bool, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_acos_cuda_bool, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_atan_cuda_int8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_atanh_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_atanh_cuda_int8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_cos_cuda_int32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_cos_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_cos_cuda_int8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_digamma_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_digamma_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_div_no_rounding_mode_cuda_int32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_erf_cuda_bool, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_exp2_cuda_int16, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_float_power_cuda_int8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_lgamma_cuda_bool, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_lgamma_cuda_int16, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_lgamma_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_log10_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_log2_cuda_int8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_log2_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_logit_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_masked_std_cuda_int32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_masked_std_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_polygamma_polygamma_n_0_cuda_int32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_polygamma_polygamma_n_1_cuda_bool, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_polygamma_polygamma_n_3_cuda_int8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_polygamma_polygamma_n_4_cuda_int16, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_reciprocal_cuda_bool, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_reciprocal_cuda_int32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_reciprocal_cuda_int8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_rsqrt_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_sin_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_sinc_cuda_int64, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_special_chebyshev_polynomial_u_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_special_hermite_polynomial_h_cuda_int32, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_special_hermite_polynomial_he_cuda_bool, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_special_laguerre_polynomial_l_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_special_shifted_chebyshev_polynomial_t_cuda_bool, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_special_shifted_chebyshev_polynomial_u_cuda_bool, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_special_shifted_chebyshev_polynomial_w_cuda_int16, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_tan_cuda_int8, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_tanh_cuda_int16, test/test_ops.py::TestCommonCUDA::test_promotes_int_to_float_true_divide_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_bfloat16_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_bool_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_bool_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_cdouble_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_cdouble_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_cdouble_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_cdouble_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_cfloat_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_cfloat_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_chalf_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_double_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_double_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_half_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_long_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs__conversions_long_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_acos_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_acosh_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_add_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_addcdiv_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_addcmul_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_addcmul_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_alias_copy_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_alias_copy_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_all_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_all_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_amin_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_any_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_as_strided_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_as_strided_scatter_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_asin_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_asin_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_asinh_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_asinh_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_asinh_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_asinh_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atan2_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atan_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atleast_2d_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atleast_3d_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_atleast_3d_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_bitwise_left_shift_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_bitwise_not_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_bitwise_or_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_bitwise_xor_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_broadcast_tensors_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_broadcast_to_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_broadcast_to_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_bucketize_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_bucketize_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_cat_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_cat_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_clone_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_clone_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_clone_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_column_stack_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_conj_physical_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_constant_pad_nd_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_constant_pad_nd_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_cosh_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_cumprod_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_cumsum_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_diag_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_diag_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_diag_embed_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_diag_embed_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_diagonal_copy_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_diagonal_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_diagonal_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_digamma_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_div_floor_rounding_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_div_floor_rounding_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_div_no_rounding_mode_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_div_no_rounding_mode_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_div_trunc_rounding_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_dsplit_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_dsplit_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_dstack_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_dstack_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_empty_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_empty_like_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_empty_like_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_empty_strided_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_eq_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_equal_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_equal_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_erf_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_erf_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_erfinv_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_exp2_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_exp2_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_exp_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_exp_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_expand_as_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_expand_as_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_expand_as_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_expand_as_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_expand_copy_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_expm1_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_expm1_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_expm1_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_eye_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_fft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_fftn_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_fftn_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_fftshift_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_hfft2_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_hfft2_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_hfft_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_hfftn_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_ifftshift_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_ifftshift_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_ifftshift_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_ihfftn_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_irfft_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_irfft_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fft_rfft2_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fill_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_flip_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_flip_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fliplr_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_float_power_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_float_power_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_floor_divide_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fmax_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fmax_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fmin_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_fmod_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_frac_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_hsplit_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_hsplit_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_hstack_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_hstack_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_hypot_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_hypot_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_i0_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_index_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_index_fill_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_index_fill_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isfinite_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isnan_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isnan_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isposinf_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isposinf_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isreal_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isreal_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_isreal_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_istft_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_le_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_le_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_linalg_diagonal_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_linalg_matrix_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_linalg_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_linspace_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log1p_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log2_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log2_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log_softmax_with_dtype_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_log_softmax_with_dtype_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logaddexp_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logical_and_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logical_and_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logical_not_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logical_not_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logical_not_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logical_not_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logical_xor_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logsumexp_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_logsumexp_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_lt_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_lt_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_masked_fill_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_mean_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_mean_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_meshgrid_list_of_tensors_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_movedim_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_movedim_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_mul_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_mul_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nan_to_num_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nan_to_num_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nan_to_num_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_narrow_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_narrow_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_native_layer_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_ne_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_neg_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_neg_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_neg_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_new_empty_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_new_empty_strided_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_new_zeros_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_alpha_dropout_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_channel_shuffle_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_elu_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_gelu_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_glu_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_hinge_embedding_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_log_softmax_with_dtype_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_margin_ranking_loss_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_nll_loss_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_pairwise_distance_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_pairwise_distance_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_pixel_shuffle_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_pixel_unshuffle_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_relu6_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_softmax_with_dtype_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_softmin_with_dtype_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_threshold_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_triplet_margin_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_triplet_margin_loss_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_triplet_margin_loss_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_nn_functional_triplet_margin_loss_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_normal__in_place_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_normal_number_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_ones_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_permute_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_permute_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_positive_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_positive_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_pow_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_prod_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_prod_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_rad2deg_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_real_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_real_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_reciprocal_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_reciprocal_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_remainder_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_remainder_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_repeat_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_reshape_as_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_reshape_as_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_rot90_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_rsqrt_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_rsqrt_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_rsub_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sgn_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sgn_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sign_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sinc_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sinc_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_softmax_with_dtype_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_bessel_j1_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_entr_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_entr_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_i0e_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_i1_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_i1e_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_log_ndtr_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_log_ndtr_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_multigammaln_mvlgamma_p_1_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_multigammaln_mvlgamma_p_3_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_multigammaln_mvlgamma_p_3_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_multigammaln_mvlgamma_p_3_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_multigammaln_mvlgamma_p_5_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_softmax_with_dtype_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_special_zeta_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sqrt_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sqrt_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sqrt_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_squeeze_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_squeeze_multiple_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_stack_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_std_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sub_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sub_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_sum_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_t_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_t_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_take_along_dim_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_tan_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_tan_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_tan_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_tan_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_tan_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_tanh_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_tanh_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_tensor_split_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_tensor_split_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_tensor_split_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_trace_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_transpose_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_transpose_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_tril_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unbind_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unflatten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unfold_copy_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unfold_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unsqueeze_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_unsqueeze_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_var_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_var_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_vdot_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_view_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_vsplit_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_vsplit_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_vstack_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref__refs_zeros_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_atan2_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_cauchy_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_eq_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_fft_hfft2_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_fft_hfftn_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_fmax_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_igamma_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_logical_xor_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_logspace_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_masked_fill_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_nn_functional_gelu_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_nn_functional_hardtanh_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_nn_functional_poisson_nll_loss_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_reshape_as_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_special_zeta_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_t_copy_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_view_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_errors__refs_where_cuda, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_T_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_byte_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_byte_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_cfloat_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_chalf_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_chalf_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_char_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_char_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_complex_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_complex_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_float_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_int_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs__conversions_int_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_acos_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_addcmul_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_addcmul_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_addr_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_addr_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_alias_copy_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_all_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_all_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_allclose_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_arange_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_arange_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_as_strided_copy_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_as_strided_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_asin_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_asin_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_asinh_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_asinh_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_atan_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_atan_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_atan_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_atleast_1d_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_bitwise_not_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_bitwise_right_shift_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_broadcast_tensors_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_broadcast_tensors_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_broadcast_to_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_cat_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_cauchy_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_chunk_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_clamp_max_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_clamp_max_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_clamp_max_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_clone_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_clone_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_column_stack_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_conj_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_conj_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_conj_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_conj_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_conj_physical_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_conj_physical_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_constant_pad_nd_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_contiguous_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_copysign_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_cosh_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_count_nonzero_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_cumsum_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_cumsum_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_deg2rad_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_deg2rad_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_diag_embed_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_diag_embed_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_diag_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_diag_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_diagonal_copy_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_diagonal_copy_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_diagonal_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_digamma_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_div_floor_rounding_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_div_no_rounding_mode_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_div_no_rounding_mode_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_dot_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_dot_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_dsplit_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_dstack_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_empty_like_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_empty_like_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_empty_strided_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_eq_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_equal_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_erfc_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_erfc_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_erfc_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_erfinv_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_exp2_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_exp2_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_expand_as_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_expand_copy_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_expand_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_expm1_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_expm1_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_expm1_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_exponential_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_eye_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_fft2_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_fft_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_fftn_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_fftn_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_fftn_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_hfft2_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_ifft2_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_irfft_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_irfftn_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_irfftn_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_rfft2_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_rfft_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fft_rfft_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fill_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fill_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_flip_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fliplr_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fliplr_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fliplr_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fliplr_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_flipud_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_float_power_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_floor_divide_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fmin_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_fmod_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_frac_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_geometric_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_gt_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_hsplit_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_hstack_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_i0_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_index_copy_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_index_fill_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_index_select_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_isclose_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_isclose_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_isinf_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_isnan_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_isnan_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_isneginf_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_isneginf_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_isposinf_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_isposinf_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_isreal_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_isreal_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_item_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_item_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_item_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_le_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_lerp_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_lgamma_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_lgamma_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_linalg_diagonal_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_linalg_diagonal_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_linalg_norm_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_linalg_svdvals_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_linalg_vecdot_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_linalg_vector_norm_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_linspace_tensor_overload_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_log1p_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_log1p_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_log2_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_log2_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_log2_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_log_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_log_softmax_with_dtype_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_log_softmax_with_dtype_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logaddexp2_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logaddexp2_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logical_and_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logical_and_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logical_or_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logical_xor_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logspace_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_logspace_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_lt_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_maximum_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_meshgrid_list_of_tensors_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_meshgrid_variadic_tensors_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_movedim_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_movedim_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_mul_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_narrow_copy_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_narrow_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_narrow_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_ne_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_ne_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_new_empty_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_new_empty_strided_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_new_empty_strided_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_new_full_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_new_full_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_new_zeros_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_new_zeros_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nextafter_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nextafter_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_alpha_dropout_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_celu_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_channel_shuffle_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_channel_shuffle_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_elu_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_gelu_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_gelu_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_glu_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_l1_loss_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_layer_norm_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_layer_norm_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_leaky_relu_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_leaky_relu_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_margin_ranking_loss_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_nll_loss_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_pixel_shuffle_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_pixel_unshuffle_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_pixel_unshuffle_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_relu6_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_selu_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_smooth_l1_loss_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_softmax_with_dtype_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_softmin_with_dtype_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_nn_functional_softplus_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_norm_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_normal__in_place_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_normal__in_place_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_permute_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_permute_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_positive_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_positive_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_prod_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_rad2deg_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_ravel_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_reciprocal_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_repeat_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_roll_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_rot90_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_rot90_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_rsub_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_select_scatter_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sgn_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sgn_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sgn_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sigmoid_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sigmoid_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sigmoid_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_signbit_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_signbit_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sin_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sin_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sin_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sinc_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sinc_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sinh_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sinh_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_softmax_with_dtype_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_entr_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_erfcx_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_i0e_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_i1_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_log_ndtr_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_multigammaln_mvlgamma_p_1_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_multigammaln_mvlgamma_p_3_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_multigammaln_mvlgamma_p_3_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_ndtr_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_ndtr_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_softmax_with_dtype_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_spherical_bessel_j0_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_zeta_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_special_zeta_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sqrt_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_square_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_squeeze_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_squeeze_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_squeeze_multiple_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_squeeze_multiple_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_stack_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_stack_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_stft_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sub_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sum_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sum_to_size_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_sum_to_size_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_t_copy_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_t_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_take_along_dim_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_tan_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_tan_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_tanh_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_tensor_split_executor_aten_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_to_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_transpose_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_tril_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_tril_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_tril_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_triu_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_trunc_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unbind_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unflatten_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unflatten_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unfold_copy_executor_aten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unfold_copy_executor_aten_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unfold_copy_executor_aten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unsqueeze_copy_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_unsqueeze_executor_aten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_var_mean_executor_aten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_view_copy_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_view_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_view_executor_aten_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_view_executor_aten_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_vsplit_executor_aten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_vsplit_executor_aten_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_vstack_executor_aten_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_vstack_executor_aten_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_executor__refs_vstack_executor_aten_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_bfloat16_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_bool_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_byte_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_cdouble_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_cdouble_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_cfloat_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_cfloat_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_cfloat_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_chalf_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_chalf_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_char_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_char_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_complex_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_float_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_float_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_half_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_int_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_int_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_short_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs__conversions_short_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_abs_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_abs_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_abs_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_acos_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_acosh_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_acosh_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_acosh_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_add_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_add_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_add_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_addcmul_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_addr_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_alias_copy_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_all_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_all_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_amin_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_amin_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_amin_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_any_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_arange_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_as_strided_copy_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_as_strided_scatter_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_as_strided_scatter_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_asin_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_asin_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atan2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atanh_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atanh_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atleast_1d_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atleast_1d_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atleast_1d_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atleast_1d_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atleast_1d_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atleast_2d_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atleast_3d_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_atleast_3d_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_bitwise_and_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_bitwise_left_shift_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_bitwise_right_shift_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_block_diag_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_broadcast_tensors_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_broadcast_to_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_broadcast_to_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_bucketize_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_ceil_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_chunk_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_chunk_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_chunk_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_chunk_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_clamp_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_clamp_max_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_clamp_max_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_clamp_min_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_clamp_min_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_clone_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_column_stack_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_conj_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_conj_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_conj_physical_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_conj_physical_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_conj_physical_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_conj_physical_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_constant_pad_nd_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_contiguous_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_copysign_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_cosh_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_count_nonzero_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_count_nonzero_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_cumprod_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_deg2rad_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_deg2rad_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_diag_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_digamma_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_div_floor_rounding_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_div_floor_rounding_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_div_no_rounding_mode_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_div_trunc_rounding_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_div_trunc_rounding_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_dot_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_dsplit_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_dsplit_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_dstack_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_dstack_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_empty_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_empty_like_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_empty_like_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_eq_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_erf_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_exp_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_expand_as_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_expand_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_expand_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_expand_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_expm1_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_fft2_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_hfft_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_ifft2_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_ifft_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_ifftn_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_ifftshift_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_irfft2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_irfft_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_irfft_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_irfft_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_irfftn_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_irfftn_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_irfftn_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_rfft_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_rfftn_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fft_rfftn_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fill_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fill_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_flatten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_flatten_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_flip_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_fliplr_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_flipud_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_flipud_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_float_power_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_frac_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_gcd_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_ge_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_geometric_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_geometric_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_geometric_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_gt_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_gt_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_heaviside_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_heaviside_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_hstack_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_i0_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_i0_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_imag_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_index_copy_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_index_copy_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_index_select_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_isinf_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_isinf_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_isnan_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_isneginf_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_isneginf_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_isposinf_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_isreal_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_lcm_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_le_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_linalg_cross_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_linalg_diagonal_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_linalg_diagonal_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_linalg_matrix_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_linalg_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_linalg_svdvals_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_linalg_vecdot_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_linalg_vector_norm_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_linalg_vector_norm_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_linspace_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_linspace_tensor_overload_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_log10_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_log1p_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_log1p_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_log2_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_log_softmax_with_dtype_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logaddexp_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logical_and_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logical_and_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logical_or_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logical_or_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logical_xor_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logspace_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logspace_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logspace_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_logsumexp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_lt_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_masked_fill_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_maximum_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_mean_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_mean_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_meshgrid_list_of_tensors_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_meshgrid_list_of_tensors_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_meshgrid_list_of_tensors_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_meshgrid_list_of_tensors_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_meshgrid_variadic_tensors_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_meshgrid_variadic_tensors_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_meshgrid_variadic_tensors_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_minimum_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_movedim_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_mul_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_narrow_copy_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_narrow_copy_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_narrow_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_ne_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_neg_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_neg_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_neg_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_new_empty_strided_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_new_empty_strided_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_new_empty_strided_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_new_full_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_new_ones_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_new_ones_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nextafter_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_celu_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_channel_shuffle_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_channel_shuffle_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_gelu_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_group_norm_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_hardtanh_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_hinge_embedding_loss_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_huber_loss_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_layer_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_margin_ranking_loss_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_nll_loss_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_pixel_shuffle_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_pixel_shuffle_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_pixel_shuffle_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_pixel_unshuffle_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_poisson_nll_loss_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_relu6_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_smooth_l1_loss_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_smooth_l1_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_softmax_with_dtype_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_softmin_with_dtype_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_softmin_with_dtype_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_softplus_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_tanhshrink_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_tanhshrink_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_triplet_margin_loss_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_nn_functional_triplet_margin_loss_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_pow_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_pow_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_pow_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_rad2deg_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_rad2deg_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_randn_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_ravel_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_real_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_real_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_real_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_reciprocal_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_remainder_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_remainder_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_renorm_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_repeat_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_repeat_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_repeat_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_repeat_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_reshape_as_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_reshape_as_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_reshape_as_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_reshape_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_roll_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_roll_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_rot90_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_rot90_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_rsqrt_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sgn_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sigmoid_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sigmoid_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sigmoid_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_signbit_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sinc_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sinc_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sinc_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sinc_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sinh_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sinh_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_bessel_j1_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_i1_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_log_softmax_with_dtype_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_multigammaln_mvlgamma_p_1_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_multigammaln_mvlgamma_p_3_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_softmax_with_dtype_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_softmax_with_dtype_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_softmax_with_dtype_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_spherical_bessel_j0_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_spherical_bessel_j0_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_xlog1py_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_special_zeta_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_square_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_squeeze_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_squeeze_multiple_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_stack_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_stack_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_std_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_stft_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sum_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sum_to_size_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_sum_to_size_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_t_copy_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_t_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_take_along_dim_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_take_along_dim_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_tan_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_tan_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_tensor_split_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_transpose_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_transpose_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_tril_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_triu_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_true_divide_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_trunc_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_unbind_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_unfold_copy_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_unfold_copy_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_unsqueeze_copy_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_unsqueeze_copy_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_unsqueeze_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_var_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_vdot_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_vdot_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_view_as_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_view_copy_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_vsplit_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_vsplit_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_vstack_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_where_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_where_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_xlogy_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_xlogy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_meta__refs_zeros_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_bfloat16_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_char_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_char_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_complex_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_complex_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_double_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_double_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs__conversions_long_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_abs_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_acos_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_add_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_alias_copy_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_alias_copy_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_alias_copy_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_allclose_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_amax_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_amax_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_amin_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_amin_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_any_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_arange_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_arange_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_as_strided_copy_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_as_strided_partial_views_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_as_strided_partial_views_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_as_strided_scatter_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_asin_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_atanh_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_atleast_2d_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_atleast_2d_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_atleast_3d_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_bitwise_and_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_bitwise_xor_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_broadcast_tensors_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_broadcast_tensors_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_broadcast_to_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_cat_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_cat_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_cauchy_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_ceil_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_clamp_min_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_clamp_min_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_column_stack_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_conj_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_conj_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_conj_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_constant_pad_nd_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_contiguous_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_cosh_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_cosh_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_cosh_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_count_nonzero_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_count_nonzero_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_cumsum_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_deg2rad_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_deg2rad_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_diag_embed_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_diagonal_copy_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_diagonal_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_diagonal_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_diagonal_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_diagonal_scatter_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_diagonal_scatter_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_digamma_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_div_trunc_rounding_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_div_trunc_rounding_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_div_trunc_rounding_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_dot_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_dsplit_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_dsplit_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_dsplit_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_dstack_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_empty_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_empty_like_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_empty_strided_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_eq_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_equal_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_equal_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_erf_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_erf_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_erf_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_exp2_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_exp_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_exp_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_expand_as_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_expand_copy_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_expand_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_expand_copy_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_expand_copy_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_expm1_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_exponential_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_eye_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_fft_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_hfft_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_hfftn_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_hfftn_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_hfftn_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_hfftn_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_ifft2_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_ifft_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_ihfft2_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_irfft_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_irfft_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_irfftn_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_rfft2_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fft_rfft_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_flatten_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_flatten_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_flip_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_flip_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_flipud_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_flipud_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_floor_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_floor_divide_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_fmin_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_frexp_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_ge_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_geometric_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_heaviside_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_heaviside_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_hsplit_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_i0_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_igamma_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_igammac_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_index_copy_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_index_copy_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_index_select_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_isfinite_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_isfinite_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_isnan_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_isnan_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_isposinf_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_isreal_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_le_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_lerp_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_lerp_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_lgamma_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_lgamma_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_linalg_cross_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_linalg_matrix_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_linalg_norm_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_linalg_norm_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_linspace_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_linspace_tensor_overload_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_log10_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_log1p_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_log2_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_log_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_log_normal_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logaddexp2_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logaddexp_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logical_xor_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logical_xor_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logical_xor_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logspace_tensor_overload_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_logspace_tensor_overload_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_masked_fill_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_masked_fill_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_maximum_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_meshgrid_list_of_tensors_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_meshgrid_list_of_tensors_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_meshgrid_variadic_tensors_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_mul_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_mul_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_narrow_copy_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_ne_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_ne_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_ne_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_neg_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_neg_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_neg_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_neg_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_empty_strided_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_empty_strided_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_empty_strided_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_full_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_full_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_ones_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_ones_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_zeros_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_new_zeros_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_celu_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_channel_shuffle_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_dropout_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_elu_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_elu_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_hinge_embedding_loss_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_hinge_embedding_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_nll_loss_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_nll_loss_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_pixel_unshuffle_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_poisson_nll_loss_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_prelu_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_relu6_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_selu_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_softmax_with_dtype_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_softmax_with_dtype_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_softmax_with_dtype_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_softmin_with_dtype_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_softmin_with_dtype_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_softmin_with_dtype_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_softmin_with_dtype_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_tanhshrink_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_tanhshrink_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_threshold_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_nn_functional_triplet_margin_loss_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_ones_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_ones_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_ones_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_permute_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_permute_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_permute_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_pow_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_pow_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_pow_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_pow_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_rad2deg_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_ravel_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_real_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_repeat_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_repeat_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_reshape_as_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_reshape_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_reshape_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_roll_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_rot90_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_rot90_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_rot90_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_rsub_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sigmoid_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sigmoid_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sign_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_signbit_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_signbit_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_signbit_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sin_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sin_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sinc_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sinh_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_softmax_with_dtype_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_softmax_with_dtype_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_softmax_with_dtype_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_entr_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_entr_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_erfcx_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_erfcx_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_i0e_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_i1_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_i1_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_i1e_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_log_softmax_with_dtype_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_log_softmax_with_dtype_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_multigammaln_mvlgamma_p_1_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_multigammaln_mvlgamma_p_3_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_multigammaln_mvlgamma_p_3_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_multigammaln_mvlgamma_p_5_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_ndtr_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_ndtri_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_spherical_bessel_j0_cuda_bool, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_xlog1py_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_special_zeta_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sqrt_cuda_int32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_square_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_squeeze_multiple_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_stack_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_stack_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_std_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_std_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_stft_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sub_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sub_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sum_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_sum_to_size_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_t_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_t_cuda_int16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_t_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_tan_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_tanh_cuda_float16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_tensor_split_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_tensor_split_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_to_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_to_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_trace_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_trace_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_transpose_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_transpose_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_true_divide_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_true_divide_cuda_int8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_unflatten_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_unflatten_cuda_float32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_unfold_copy_cuda_int64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_unfold_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_unsqueeze_cuda_bfloat16, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_unsqueeze_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_var_mean_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_vdot_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_view_as_cuda_float64, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_view_as_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_view_copy_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_view_cuda_complex32, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_view_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_vsplit_cuda_complex128, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_vstack_cuda_uint8, test/test_ops.py::TestCommonCUDA::test_python_ref_torch_fallback__refs_xlogy_cuda_int64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager___rmod___cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager__unsafe_masked_index_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_acos_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_addbmm_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_as_strided_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_bfloat16_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_broadcast_tensors_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_chalf_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_char_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_cholesky_inverse_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_cholesky_solve_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_clone_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_conj_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_count_nonzero_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_count_nonzero_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_diagonal_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_div_no_rounding_mode_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_div_no_rounding_mode_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_dstack_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_empty_like_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_empty_permuted_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_exponential_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_fft_ihfft_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_fft_irfft_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_fmax_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_index_reduce_mean_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_index_reduce_prod_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_isreal_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_jiterator_binary_return_by_ref_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_kron_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_det_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_lstsq_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_pinv_singular_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_solve_triangular_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_tensorinv_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_vander_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_log2_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_log_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_logcumsumexp_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_logcumsumexp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_logit_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_masked_fill_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_masked_logaddexp_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_masked_softmax_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_median_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_mul_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nanmean_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_new_empty_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_new_ones_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nextafter_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_adaptive_avg_pool3d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_avg_pool3d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_conv_transpose1d_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_conv_transpose1d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_cross_entropy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_elu_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_fractional_max_pool2d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_hardshrink_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_linear_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_max_unpool1d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_max_unpool3d_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_mish_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_normalize_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_pad_constant_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_pad_replicate_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_nn_functional_triplet_margin_with_distance_loss_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_ones_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_quantile_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_rand_like_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_randint_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_reciprocal_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_reshape_as_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_resize_as__cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_rsqrt_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_scalar_tensor_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_searchsorted_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_signal_windows_cosine_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_signal_windows_gaussian_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_sinc_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_special_chebyshev_polynomial_u_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_special_chebyshev_polynomial_w_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_special_hermite_polynomial_he_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_special_i0e_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_special_modified_bessel_i0_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_special_modified_bessel_k0_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_split_with_sizes_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_split_with_sizes_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_std_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_std_mean_unbiased_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_std_mean_unbiased_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_std_unbiased_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_tile_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_to_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_to_sparse_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_transpose_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_triangular_solve_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_unsqueeze_copy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_var_cuda_complex64, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_var_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_xlogy_cuda_float32, test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_zeros_like_cuda_complex64, test/test_ops.py::TestCompositeComplianceCUDA::test_backward___rmatmul___cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward__segment_reduce_lengths_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_addbmm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_addr_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_as_strided_scatter_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_atan_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_bfloat16_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_cholesky_inverse_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_cholesky_solve_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_conj_physical_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_corrcoef_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_cov_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_cummin_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_expm1_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_fft_hfftn_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_fft_ifft2_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_fft_ihfft_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_index_reduce_mean_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_linalg_det_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_linalg_norm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_linalg_slogdet_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_linalg_svdvals_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_log_softmax_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_masked_cumprod_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_masked_logsumexp_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_masked_softmax_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_masked_softmin_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_max_reduction_no_dim_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_meshgrid_variadic_tensors_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_min_reduction_no_dim_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_mm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_movedim_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_mul_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_native_layer_norm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_adaptive_max_pool3d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_hardsigmoid_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_interpolate_trilinear_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_max_pool2d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_max_unpool1d_grad_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_multi_head_attention_forward_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_multi_margin_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_nll_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_relu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_softsign_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_nn_functional_threshold_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_outer_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_polygamma_polygamma_n_3_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_roll_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_select_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_sort_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_special_i1_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_special_log_ndtr_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_split_with_sizes_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_stack_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_t_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_t_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_tensordot_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_var_unbiased_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_backward_vstack_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input___rmul___cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input__segment_reduce_offsets_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_abs_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_addcdiv_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_amin_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_as_strided_scatter_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_block_diag_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_cauchy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_char_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_clamp_max_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_clamp_min_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_clone_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_column_stack_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_cosh_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_cumsum_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_deg2rad_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_dot_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_eq_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_erfinv_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_fft_fftn_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_float_power_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_fmin_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_index_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_index_put_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_isin_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_jiterator_2inputs_2outputs_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_linalg_det_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_linalg_householder_product_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_linalg_ldl_factor_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_linalg_ldl_factor_ex_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_lu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_masked_amax_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_masked_logaddexp_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_masked_mean_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_min_binary_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_new_ones_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_binary_cross_entropy_with_logits_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_conv2d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_hardsigmoid_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_nll_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_relu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_smooth_l1_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_nn_functional_softplus_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_prod_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_randint_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_special_chebyshev_polynomial_u_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_special_modified_bessel_i1_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_special_scaled_modified_bessel_k1_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_squeeze_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_std_mean_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_std_unbiased_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_sum_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_tile_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_torch_ops_aten__safe_softmax_default_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_triu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_unsafe_chunk_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_var_unbiased_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_view_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_vstack_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_cow_input_xlogy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad___rmod___cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_amin_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_argsort_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_as_strided_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_atanh_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_cauchy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_cdist_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_clamp_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_complex_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_cummax_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_div_trunc_rounding_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_einsum_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_erf_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_expand_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_fft_ihfft_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_frexp_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_hstack_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_isinf_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_linalg_cholesky_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_linalg_det_singular_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_linalg_householder_product_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_linalg_lu_factor_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_linalg_qr_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_linalg_vecdot_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_log2_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_logical_or_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_logical_xor_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_masked_fill_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_mvlgamma_mvlgamma_p_1_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nansum_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_adaptive_avg_pool1d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_adaptive_max_pool2d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_avg_pool2d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_bilinear_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_binary_cross_entropy_with_logits_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_conv2d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_gaussian_nll_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_max_unpool2d_grad_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_max_unpool3d_grad_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_relu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_nn_functional_selu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_polygamma_polygamma_n_0_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_repeat_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_scatter_reduce_mean_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_signal_windows_kaiser_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_sin_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_softmax_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_special_chebyshev_polynomial_t_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_special_hermite_polynomial_he_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_special_i0e_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_special_shifted_chebyshev_polynomial_u_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_take_along_dim_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_forward_ad_view_as_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator___radd___cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator___rmul___cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_add_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_alias_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_all_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_as_strided_partial_views_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_broadcast_to_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_count_nonzero_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_cumprod_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_diag_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_eq_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_expm1_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_fft_hfft2_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_fft_irfft_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_fft_rfft2_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_fill_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_geometric_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_index_put_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_index_reduce_mean_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_le_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linalg_inv_ex_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linalg_ldl_factor_ex_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linalg_lstsq_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linalg_lstsq_grad_oriented_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linalg_norm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_linalg_tensorsolve_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_log_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_logaddexp2_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_logspace_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_mH_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_movedim_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_multinomial_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_narrow_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_elu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_fractional_max_pool3d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_grid_sample_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_hardshrink_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_hardswish_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_interpolate_nearest_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_kl_div_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nn_functional_normalize_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_nonzero_static_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_prod_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_put_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_reciprocal_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_round_decimals_3_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_round_decimals_neg_3_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_special_airy_ai_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_special_chebyshev_polynomial_v_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_special_modified_bessel_k0_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_special_shifted_chebyshev_polynomial_t_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_std_unbiased_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_tanh_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_tile_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_trace_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_trapz_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_unbind_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_var_mean_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_var_unbiased_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_operator_view_as_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay___getitem___cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay__unsafe_masked_index_put_accumulate_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_alias_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_as_strided_scatter_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_bfloat16_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_bmm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_cfloat_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_chalf_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_cholesky_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_cross_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_cummin_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_dot_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_empty_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_expand_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_expm1_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_exponential_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_fft_fft2_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_fft_ifftn_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_fft_irfft_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_fft_rfftn_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_flip_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_float_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_floor_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_half_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_heaviside_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_i0_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_item_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_kthvalue_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_linalg_inv_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_linalg_lu_factor_ex_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_linspace_tensor_overload_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_log2_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_log_softmax_with_dtype_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_logit_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_lu_unpack_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_masked_argmin_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_masked_cumsum_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_masked_logsumexp_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_masked_norm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_movedim_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_mv_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nanmedian_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nansum_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_avg_pool3d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_conv_transpose2d_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_ctc_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_dropout_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_feature_alpha_dropout_with_train_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_gaussian_nll_loss_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_interpolate_nearest_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_pad_circular_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_prelu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_relu6_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_nn_functional_relu_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_normal_in_place_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_pca_lowrank_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_quantile_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_reciprocal_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_round_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_round_decimals_0_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_scalar_tensor_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_scatter_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_sgn_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_short_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_sinc_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_sparse_mm_reduce_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_sparse_sampled_addmm_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_special_bessel_y0_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_special_chebyshev_polynomial_w_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_special_hermite_polynomial_he_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_special_spherical_bessel_j0_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_split_with_sizes_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_svd_lowrank_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_tensor_split_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_trunc_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_unbind_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_unfold_copy_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_unique_consecutive_cuda_float32, test/test_ops.py::TestCompositeComplianceCUDA::test_view_replay_var_mean_cuda_float32, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs__conversions_double_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_addcdiv_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_all_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_any_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_atleast_3d_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_empty_like_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_empty_strided_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_exp2_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_fft_ifftn_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_fft_ifftshift_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_flatten_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_lerp_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_new_full_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_new_zeros_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_nn_functional_softmin_with_dtype_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_reshape_as_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_sqrt_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_squeeze_multiple_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_stft_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_unflatten_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__refs_view_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view__unsafe_masked_index_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_addmm_decomposed_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_all_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_angle_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_byte_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_clone_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_corrcoef_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_diagonal_copy_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_exp2_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_fft_hfft_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_fft_ifft_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_fliplr_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_half_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_isclose_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_isfinite_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_jiterator_unary_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_linalg_eigvalsh_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_linalg_matrix_norm_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_linalg_matrix_rank_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_linalg_pinv_hermitian_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_log10_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_long_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_mT_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_masked_std_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_meshgrid_variadic_tensors_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_new_empty_strided_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_nn_functional_l1_loss_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_nonzero_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_nonzero_static_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_ones_like_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_pinverse_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_sinh_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_sum_to_size_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_tensor_split_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_trapz_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_conj_view_view_as_cuda_complex64, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_H_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_T_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view___getitem___cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view___rdiv___cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view___rpow___cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_T_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs__conversions_int_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_allclose_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_chunk_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_conj_physical_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_count_nonzero_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_equal_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_fft_fftn_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_fft_fftshift_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_fft_ifftshift_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_log_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_movedim_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_renorm_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_roll_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_sgn_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_sum_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_sum_to_size_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_take_along_dim_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_to_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_true_divide_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_unsqueeze_copy_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view__refs_vsplit_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_abs_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_acos_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_addmm_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_as_strided_copy_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_as_strided_partial_views_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_asin_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_exp_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_expand_as_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_expm1_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_fft_fft_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_fft_hfftn_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_fft_irfftn_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_flipud_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_float_power_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_index_copy_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_jiterator_2inputs_2outputs_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_linalg_eig_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_linalg_lstsq_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_linalg_matrix_power_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_linalg_svd_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_logdet_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_lu_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_masked_cumsum_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_matrix_exp_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_meshgrid_variadic_tensors_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_nn_functional_conv_transpose2d_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_nn_functional_softmin_with_dtype_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_pinverse_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_rand_like_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_rot90_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_scalar_tensor_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_split_with_sizes_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_squeeze_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_stft_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_conj_view_tan_cuda_complex128, test/test_ops.py::TestMathBitsCUDA::test_neg_view___rdiv___cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_add_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_addcdiv_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_alias_copy_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_contiguous_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_equal_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_fft_hfft_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_fft_irfft2_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_fft_rfft_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_fliplr_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_igamma_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_isclose_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_logical_not_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_maximum_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_nn_functional_channel_shuffle_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_nn_functional_layer_norm_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_nn_functional_log_softmax_with_dtype_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_nn_functional_nll_loss_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_nn_functional_relu6_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_nn_functional_softshrink_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_rad2deg_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_sign_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_sum_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_transpose_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view__refs_view_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_addcmul_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_addr_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_all_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_amin_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_atleast_1d_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_bmm_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_cov_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_diagflat_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_diagonal_scatter_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_erfc_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_expand_as_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_fft_ifftn_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_hstack_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_igammac_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_isclose_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_isinf_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_jiterator_binary_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_kthvalue_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_linalg_cholesky_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_linalg_diagonal_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_linalg_norm_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_linalg_slogdet_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_linalg_solve_ex_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_linalg_tensorsolve_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_logical_or_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_masked_logsumexp_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_masked_median_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_masked_scatter_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_masked_sum_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_matrix_exp_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_meshgrid_list_of_tensors_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_mvlgamma_mvlgamma_p_1_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_mvlgamma_mvlgamma_p_5_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nanmedian_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_native_dropout_backward_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_avg_pool1d_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_l1_loss_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_logsigmoid_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nn_functional_pdist_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_nonzero_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_qr_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_ravel_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_real_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_repeat_interleave_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_resolve_neg_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_roll_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_scatter_reduce_amax_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_short_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_signal_windows_hamming_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_slice_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_special_i1_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_special_polygamma_special_polygamma_n_0_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_special_zeta_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_sub_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_sum_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_take_along_dim_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_topk_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_unique_consecutive_cuda_float64, test/test_ops.py::TestMathBitsCUDA::test_neg_view_var_mean_unbiased_cuda_float64, test/test_ops.py::TestFakeTensorCUDA::test_fake_all_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast___rmatmul___cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast__chunk_cat_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_addcmul_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_any_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_atleast_1d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_cholesky_solve_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_clamp_max_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_div_floor_rounding_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_fft_rfft_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_fft_rfftn_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_floor_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_igamma_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_jiterator_2inputs_2outputs_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_jiterator_binary_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_le_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_linalg_householder_product_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_linalg_norm_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_linalg_pinv_singular_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_linspace_tensor_overload_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_log_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_log_softmax_with_dtype_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_logaddexp_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_logdet_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_long_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_lu_unpack_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_masked_fill_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_masked_norm_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_masked_softmin_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_matmul_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_maximum_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_meshgrid_list_of_tensors_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_minimum_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_mvlgamma_mvlgamma_p_3_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_new_zeros_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_adaptive_avg_pool1d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_alpha_dropout_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_avg_pool1d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_conv2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_elu_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_embedding_bag_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_feature_alpha_dropout_without_train_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_hardtanh_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_linear_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_multi_margin_loss_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_pairwise_distance_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_pixel_unshuffle_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_nn_functional_prelu_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_ones_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_polygamma_polygamma_n_4_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_pow_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_scatter_reduce_mean_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_searchsorted_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_signal_windows_exponential_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_signal_windows_hann_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_special_bessel_j0_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_special_i0e_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_sqrt_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_squeeze_multiple_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_svd_lowrank_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_trapezoid_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_uniform_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_unravel_index_cuda_int64, test/test_ops.py::TestFakeTensorCUDA::test_fake_autocast_view_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_bitwise_xor_cuda_int64, test/test_ops.py::TestFakeTensorCUDA::test_fake_broadcast_to_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_clamp_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_clamp_max_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_asinh_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_broadcast_to_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_cross_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_cumulative_trapezoid_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_diagflat_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_div_no_rounding_mode_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_erfc_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_exp2_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_fft_hfft_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_fft_ifft_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_index_fill_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_index_select_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_linalg_lstsq_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_linalg_norm_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_logaddexp2_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_logaddexp_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_masked_logsumexp_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_masked_mean_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_median_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_minimum_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_mm_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_mul_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_adaptive_avg_pool1d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_adaptive_avg_pool3d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_adaptive_max_pool3d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_conv_transpose2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_elu_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_embedding_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_hardtanh_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_interpolate_bilinear_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_normalize_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_pad_reflect_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_prelu_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_softsign_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_nn_functional_triplet_margin_with_distance_loss_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_norm_inf_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_rad2deg_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_reciprocal_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_sin_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_softmax_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_sub_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_tan_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_tensor_split_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_trapz_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_triangular_solve_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_unfold_copy_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_unsqueeze_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_amp_view_as_complex_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp___rdiv___cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp__segment_reduce_offsets_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_addcdiv_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_addmv_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_amax_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_as_strided_partial_views_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_atleast_2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_bernoulli_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_block_diag_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_cholesky_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_clone_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_diagonal_scatter_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_exp2_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_expand_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_fft_hfft2_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_frac_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_index_fill_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_ldexp_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_linalg_householder_product_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_linalg_solve_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_linalg_vecdot_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_linalg_vector_norm_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_logcumsumexp_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_lu_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_masked_amin_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_masked_cumsum_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_meshgrid_list_of_tensors_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_native_dropout_backward_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_adaptive_avg_pool1d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_ctc_loss_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_dropout3d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_dropout_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_elu_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_embedding_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_hardsigmoid_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_hardtanh_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_instance_norm_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_max_unpool2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_nn_functional_pad_replicate_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_polygamma_polygamma_n_0_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_polygamma_polygamma_n_1_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_rsqrt_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_scatter_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_scatter_reduce_amax_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_select_scatter_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_sinh_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_slice_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_slice_scatter_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_split_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_sub_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_tanh_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_unbind_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_unsafe_split_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_vdot_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_crossref_backward_no_amp_vsplit_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_diff_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_dsplit_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_einsum_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_erfinv_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_fft_ihfft2_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_fill_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_gather_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_imag_cuda_complex64, test/test_ops.py::TestFakeTensorCUDA::test_fake_index_reduce_amax_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_istft_cuda_complex64, test/test_ops.py::TestFakeTensorCUDA::test_fake_jiterator_4inputs_with_extra_args_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_linalg_cond_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_linalg_lu_solve_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_linalg_svd_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_log1p_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_logdet_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_masked_log_softmax_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_meshgrid_variadic_tensors_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_mvlgamma_mvlgamma_p_5_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_native_dropout_backward_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_adaptive_max_pool1d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_avg_pool2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_bilinear_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_binary_cross_entropy_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_cross_entropy_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_hardswish_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_huber_loss_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_logsigmoid_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_max_pool3d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_max_unpool3d_grad_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_pad_replicate_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_pad_replicate_negative_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_nn_functional_poisson_nll_loss_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_roll_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_sgn_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_signal_windows_bartlett_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_signal_windows_blackman_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_sparse_mm_reduce_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_special_i1_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_special_i1e_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_special_laguerre_polynomial_l_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_split_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_split_with_sizes_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_std_mean_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_stft_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_svd_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_tile_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_torch_ops_aten__efficient_attention_forward_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_torch_ops_aten__flash_attention_forward_cuda_float16, test/test_ops.py::TestFakeTensorCUDA::test_fake_trunc_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_fake_view_as_complex_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops___rpow___cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_acos_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_acosh_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_addcdiv_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_addcmul_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_asin_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_ceil_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_clamp_min_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_clone_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_contiguous_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_copysign_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_cov_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_cross_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_deg2rad_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_dist_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_empty_like_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_fft_hfft2_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_fft_ihfft_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_floor_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_istft_cuda_complex64, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_lgamma_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_linalg_eig_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_linalg_ldl_factor_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_linalg_vander_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_logical_or_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_logsumexp_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_lu_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_masked_argmax_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_masked_mean_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_masked_scatter_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_max_pool2d_with_indices_backward_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_max_reduction_with_dim_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_nextafter_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_nn_functional_conv2d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_nn_functional_gaussian_nll_loss_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_nn_functional_linear_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_nn_functional_max_pool3d_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_nn_functional_multi_margin_loss_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_nn_functional_rrelu_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_randn_like_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_real_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_renorm_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_resize__cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_round_decimals_3_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_scatter_add_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_scatter_reduce_amax_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_searchsorted_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_sgn_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_signal_windows_exponential_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_signal_windows_hann_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_sort_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_sparse_mm_reduce_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_special_erfcx_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_trace_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_var_unbiased_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_vdot_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_pointwise_ops_view_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout__refs_arange_cuda_float16, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout__refs_arange_cuda_int16, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout__refs_linspace_tensor_overload_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout__refs_ones_cuda_float16, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout__refs_zeros_cuda_float64, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout_full_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout_full_cuda_int64, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout_logspace_cuda_float32, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout_logspace_tensor_overload_cuda_complex64, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout_logspace_tensor_overload_cuda_int8, test/test_ops.py::TestFakeTensorCUDA::test_strided_layout_ones_cuda_complex128, test/test_ops.py::TestTagsCUDA::test_tags___rmatmul___cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags___ror___cuda_int64, test/test_ops.py::TestTagsCUDA::test_tags__refs__conversions_half_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_abs_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_asinh_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_chunk_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_clone_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_conj_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_copysign_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_exp_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_expand_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_fft_fftshift_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_fft_rfft2_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_fft_rfft_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_fliplr_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_frexp_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_index_add_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_index_copy_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_lt_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_movedim_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_new_empty_strided_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_nn_functional_pdist_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_nn_functional_pixel_shuffle_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_roll_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_sinc_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_stack_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_std_mean_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_sum_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_transpose_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags__refs_triu_indices_cuda_int64, test/test_ops.py::TestTagsCUDA::test_tags__refs_unfold_copy_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_abs_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_atleast_3d_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_clamp_min_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_cos_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_cummax_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_deg2rad_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_diagflat_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_double_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_erf_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_fft_hfft2_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_fft_ifftn_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_fft_ifftshift_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_flipud_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_fmin_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_gcd_cuda_int64, test/test_ops.py::TestTagsCUDA::test_tags_i0_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_index_put_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_index_reduce_mean_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_isneginf_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_jiterator_2inputs_2outputs_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_jiterator_binary_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_linalg_ldl_factor_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_linalg_ldl_solve_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_linalg_norm_subgradients_at_zero_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_log2_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nanquantile_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_new_full_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_celu_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_cosine_embedding_loss_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_ctc_loss_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_gelu_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_mse_loss_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_multilabel_soft_margin_loss_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_pad_circular_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_selu_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_nn_functional_soft_margin_loss_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_pinverse_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_polar_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_prod_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_randn_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_randn_like_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_renorm_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_scatter_add_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_select_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_select_scatter_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_short_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_special_scaled_modified_bessel_k1_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_stack_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_stft_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_to_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_trapz_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_unsafe_split_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_unsqueeze_copy_cuda_float32, test/test_ops.py::TestTagsCUDA::test_tags_zeros_like_cuda_float32 2024-08-06T21:49:19.9133620Z 2024-08-06T21:49:23.3635213Z Running inductor/test_inplacing_pass 1/1 ... [2024-08-06 21:49:23.363003] 2024-08-06T21:49:23.3638892Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_inplacing_pass.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:49:23.363456] 2024-08-06T21:49:26.8819736Z 2024-08-06T21:49:26.8820864Z inductor/test_inplacing_pass 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_inplacing_pass_1.1_5b183ea65d3b7250_.log 2024-08-06T21:49:26.8821726Z 2024-08-06T21:49:30.6866861Z Running export/test_tools 1/1 ... [2024-08-06 21:49:30.686212] 2024-08-06T21:49:30.6871172Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'export/test_tools.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:49:30.686693] 2024-08-06T21:49:35.3595563Z 2024-08-06T21:49:35.3596946Z export/test_tools 1/1 was successful, full logs can be found in artifacts with path test/test-reports/export.test_tools_1.1_6607c3bcb2b9bbdb_.log 2024-08-06T21:49:35.3598580Z Running 2 items in this shard: test/export/test_tools.py::TestExportTools::test_report_exportability_basic, test/export/test_tools.py::TestExportTools::test_report_exportability_with_issues 2024-08-06T21:49:35.3599540Z 2024-08-06T21:49:39.0180875Z Running inductor/test_torchinductor_dynamic_shapes 1/4 ... [2024-08-06 21:49:39.017631] 2024-08-06T21:49:39.0185572Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_torchinductor_dynamic_shapes.py', '-m', 'not serial', '--shard-id=1', '--num-shards=4', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:49:39.018126] 2024-08-06T21:51:06.7044711Z 2024-08-06T21:51:06.7049072Z functorch/test_vmap 1/1 was successful, full logs can be found in artifacts with path test/test-reports/functorch.test_vmap_1.1_d471fca716b413f4_.log 2024-08-06T21:51:06.8191647Z Running 2113 items in this shard: test/functorch/test_vmap.py::TestVmapAPI::test_accepts_nested_inputs, test/functorch/test_vmap.py::TestVmapAPI::test_backward_unsupported_interaction, test/functorch/test_vmap.py::TestVmapAPI::test_batch_rule_does_not_need_to_handle_no_batched_input, test/functorch/test_vmap.py::TestVmapAPI::test_batched_gradient_basic, test/functorch/test_vmap.py::TestVmapAPI::test_checkpoint, test/functorch/test_vmap.py::TestVmapAPI::test_constant_function, test/functorch/test_vmap.py::TestVmapAPI::test_data_attribute, test/functorch/test_vmap.py::TestVmapAPI::test_data_dependent_control_flow_throws, test/functorch/test_vmap.py::TestVmapAPI::test_decomposition_under_python_dispatcher, test/functorch/test_vmap.py::TestVmapAPI::test_different_map_dim_size_raises, test/functorch/test_vmap.py::TestVmapAPI::test_fallback_does_not_warn_by_default, test/functorch/test_vmap.py::TestVmapAPI::test_fallback_masked_fill, test/functorch/test_vmap.py::TestVmapAPI::test_fallback_multiple_returns, test/functorch/test_vmap.py::TestVmapAPI::test_fallback_warning, test/functorch/test_vmap.py::TestVmapAPI::test_fallback_warns_when_warnings_are_enabled, test/functorch/test_vmap.py::TestVmapAPI::test_fallback_with_undefined_grad, test/functorch/test_vmap.py::TestVmapAPI::test_fallback_zero_dim, test/functorch/test_vmap.py::TestVmapAPI::test_func_with_no_inputs, test/functorch/test_vmap.py::TestVmapAPI::test_func_with_no_tensors, test/functorch/test_vmap.py::TestVmapAPI::test_functools_partial, test/functorch/test_vmap.py::TestVmapAPI::test_grad_unsupported_interaction, test/functorch/test_vmap.py::TestVmapAPI::test_in_dim_not_in_tensor_err_msg, test/functorch/test_vmap.py::TestVmapAPI::test_in_dims_wrong_type_err_msg, test/functorch/test_vmap.py::TestVmapAPI::test_inplace_fallback_nary_different_levels, test/functorch/test_vmap.py::TestVmapAPI::test_inplace_fallback_nary_same_levels, test/functorch/test_vmap.py::TestVmapAPI::test_inplace_fallback_unary, test/functorch/test_vmap.py::TestVmapAPI::test_integer_in_dim_but_not_tensor_input_err_msg, test/functorch/test_vmap.py::TestVmapAPI::test_item_throws, test/functorch/test_vmap.py::TestVmapAPI::test_multiple_inputs, test/functorch/test_vmap.py::TestVmapAPI::test_multiple_out_dims, test/functorch/test_vmap.py::TestVmapAPI::test_multiple_outputs, test/functorch/test_vmap.py::TestVmapAPI::test_multiple_outputs2, test/functorch/test_vmap.py::TestVmapAPI::test_nested_negative_in_dims, test/functorch/test_vmap.py::TestVmapAPI::test_nested_non_default_in_dims, test/functorch/test_vmap.py::TestVmapAPI::test_nested_out_dims, test/functorch/test_vmap.py::TestVmapAPI::test_nested_with_diag_embed, test/functorch/test_vmap.py::TestVmapAPI::test_nested_with_different_map_dim, test/functorch/test_vmap.py::TestVmapAPI::test_nested_with_same_map_dim, test/functorch/test_vmap.py::TestVmapAPI::test_nn_module, test/functorch/test_vmap.py::TestVmapAPI::test_non_default_in_dims_out_dims, test/functorch/test_vmap.py::TestVmapAPI::test_non_tensor_output_raises, test/functorch/test_vmap.py::TestVmapAPI::test_non_zero_in_dims, test/functorch/test_vmap.py::TestVmapAPI::test_none_in_dims, test/functorch/test_vmap.py::TestVmapAPI::test_nonzero_out_dims, test/functorch/test_vmap.py::TestVmapAPI::test_noop_in_inner_vmap, test/functorch/test_vmap.py::TestVmapAPI::test_not_enough_in_dims_err_msg, test/functorch/test_vmap.py::TestVmapAPI::test_out_dim_out_of_bounds_err_msg, test/functorch/test_vmap.py::TestVmapAPI::test_out_dims_and_num_outputs_mismatch_err_msg, test/functorch/test_vmap.py::TestVmapAPI::test_out_dims_edge_case, test/functorch/test_vmap.py::TestVmapAPI::test_out_dims_must_be_int_or_collection_of_int_err_msg, test/functorch/test_vmap.py::TestVmapAPI::test_out_dims_none, test/functorch/test_vmap.py::TestVmapAPI::test_out_dims_none_tuple, test/functorch/test_vmap.py::TestVmapAPI::test_out_dims_normal_tensor, test/functorch/test_vmap.py::TestVmapAPI::test_pytree_odict_returns, test/functorch/test_vmap.py::TestVmapAPI::test_pytree_returns, test/functorch/test_vmap.py::TestVmapAPI::test_pytree_returns_broadcast_nested, test/functorch/test_vmap.py::TestVmapAPI::test_pytree_returns_broadcast_simple, test/functorch/test_vmap.py::TestVmapAPI::test_pytree_returns_outdims, test/functorch/test_vmap.py::TestVmapAPI::test_reshape_dim_into, test/functorch/test_vmap.py::TestVmapAPI::test_reshape_dim_outof, test/functorch/test_vmap.py::TestVmapAPI::test_restore_vmap_no_vmapped_inputs, test/functorch/test_vmap.py::TestVmapAPI::test_restore_vmap_pytree_input_output, test/functorch/test_vmap.py::TestVmapAPI::test_restore_vmap_unexpanded_outputs, test/functorch/test_vmap.py::TestVmapAPI::test_single_input, test/functorch/test_vmap.py::TestVmapAPI::test_unsupported_op_err_msg, test/functorch/test_vmap.py::TestVmapAPI::test_vmap_autocast_cpu, test/functorch/test_vmap.py::TestVmapAPI::test_vmap_autocast_cuda, test/functorch/test_vmap.py::TestVmapOperators::test_T_numpy, test/functorch/test_vmap.py::TestVmapOperators::test_adaptive_avg_pool2d, test/functorch/test_vmap.py::TestVmapOperators::test_argmax_dim, test/functorch/test_vmap.py::TestVmapOperators::test_arithmetic_add, test/functorch/test_vmap.py::TestVmapOperators::test_arithmetic_add_dunder, test/functorch/test_vmap.py::TestVmapOperators::test_arithmetic_div, test/functorch/test_vmap.py::TestVmapOperators::test_arithmetic_div_dunder, test/functorch/test_vmap.py::TestVmapOperators::test_arithmetic_mul, test/functorch/test_vmap.py::TestVmapOperators::test_arithmetic_mul_dunder, test/functorch/test_vmap.py::TestVmapOperators::test_arithmetic_pow, test/functorch/test_vmap.py::TestVmapOperators::test_arithmetic_pow_dunder, test/functorch/test_vmap.py::TestVmapOperators::test_arithmetic_sub, test/functorch/test_vmap.py::TestVmapOperators::test_arithmetic_sub_dunder, test/functorch/test_vmap.py::TestVmapOperators::test_as_strided, test/functorch/test_vmap.py::TestVmapOperators::test_bmm, test/functorch/test_vmap.py::TestVmapOperators::test_cat, test/functorch/test_vmap.py::TestVmapOperators::test_chunk, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_0_out_dim_0_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_0_out_dim_0_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_0_out_dim_1_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_0_out_dim_1_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_0_out_dim_2_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_0_out_dim_2_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_1_out_dim_0_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_1_out_dim_0_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_1_out_dim_1_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_1_out_dim_1_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_1_out_dim_2_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_1_out_dim_2_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_2_out_dim_0_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_2_out_dim_0_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_2_out_dim_1_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_2_out_dim_1_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_2_out_dim_2_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_chunk_vmap_in_dim_2_out_dim_2_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_clamp, test/functorch/test_vmap.py::TestVmapOperators::test_clamp_inplace_variant_clamp_max_, test/functorch/test_vmap.py::TestVmapOperators::test_clamp_inplace_variant_clamp_min_, test/functorch/test_vmap.py::TestVmapOperators::test_clamp_variant_clamp_max, test/functorch/test_vmap.py::TestVmapOperators::test_clamp_variant_clamp_min, test/functorch/test_vmap.py::TestVmapOperators::test_clone, test/functorch/test_vmap.py::TestVmapOperators::test_comparison_ops, test/functorch/test_vmap.py::TestVmapOperators::test_conj, test/functorch/test_vmap.py::TestVmapOperators::test_conj_bit, test/functorch/test_vmap.py::TestVmapOperators::test_contiguous, test/functorch/test_vmap.py::TestVmapOperators::test_conv2d, test/functorch/test_vmap.py::TestVmapOperators::test_copy_, test/functorch/test_vmap.py::TestVmapOperators::test_cross_batch_size_three, test/functorch/test_vmap.py::TestVmapOperators::test_diagonal, test/functorch/test_vmap.py::TestVmapOperators::test_dot, test/functorch/test_vmap.py::TestVmapOperators::test_expand_as, test/functorch/test_vmap.py::TestVmapOperators::test_fill_and_zero_inplace, test/functorch/test_vmap.py::TestVmapOperators::test_imag, test/functorch/test_vmap.py::TestVmapOperators::test_is_complex, test/functorch/test_vmap.py::TestVmapOperators::test_is_contiguous, test/functorch/test_vmap.py::TestVmapOperators::test_is_floating_point, test/functorch/test_vmap.py::TestVmapOperators::test_mean, test/functorch/test_vmap.py::TestVmapOperators::test_mean_dim, test/functorch/test_vmap.py::TestVmapOperators::test_mm, test/functorch/test_vmap.py::TestVmapOperators::test_mode_key, test/functorch/test_vmap.py::TestVmapOperators::test_movedim, test/functorch/test_vmap.py::TestVmapOperators::test_mv, test/functorch/test_vmap.py::TestVmapOperators::test_narrow, test/functorch/test_vmap.py::TestVmapOperators::test_new_empty, test/functorch/test_vmap.py::TestVmapOperators::test_new_empty_strided, test/functorch/test_vmap.py::TestVmapOperators::test_new_zeros, test/functorch/test_vmap.py::TestVmapOperators::test_nll_loss, test/functorch/test_vmap.py::TestVmapOperators::test_one_hot, test/functorch/test_vmap.py::TestVmapOperators::test_real, test/functorch/test_vmap.py::TestVmapOperators::test_repeat, test/functorch/test_vmap.py::TestVmapOperators::test_reshape, test/functorch/test_vmap.py::TestVmapOperators::test_reshape_as, test/functorch/test_vmap.py::TestVmapOperators::test_result_type, test/functorch/test_vmap.py::TestVmapOperators::test_roll_no_dims, test/functorch/test_vmap.py::TestVmapOperators::test_select, test/functorch/test_vmap.py::TestVmapOperators::test_silu_backward, test/functorch/test_vmap.py::TestVmapOperators::test_slice, test/functorch/test_vmap.py::TestVmapOperators::test_slogdet, test/functorch/test_vmap.py::TestVmapOperators::test_split, test/functorch/test_vmap.py::TestVmapOperators::test_squeeze, test/functorch/test_vmap.py::TestVmapOperators::test_stack, test/functorch/test_vmap.py::TestVmapOperators::test_stride, test/functorch/test_vmap.py::TestVmapOperators::test_sum, test/functorch/test_vmap.py::TestVmapOperators::test_sum_dim, test/functorch/test_vmap.py::TestVmapOperators::test_t, test/functorch/test_vmap.py::TestVmapOperators::test_tensor_split, test/functorch/test_vmap.py::TestVmapOperators::test_to, test/functorch/test_vmap.py::TestVmapOperators::test_trace, test/functorch/test_vmap.py::TestVmapOperators::test_transpose, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_abs, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_acos, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_asin, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_atan, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_ceil, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_cos, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_cosh, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_digamma, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_exp, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_expm1, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_floor, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_frac, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_lgamma, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_log, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_log10, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_log1p, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_log2, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_neg, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_reciprocal, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_relu, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_round, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_rsqrt, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_sigmoid, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_sign, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_sin, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_sinh, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_sqrt, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_tan, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_tanh, test/functorch/test_vmap.py::TestVmapOperators::test_unary_pointwise_trunc, test/functorch/test_vmap.py::TestVmapOperators::test_unbind, test/functorch/test_vmap.py::TestVmapOperators::test_unfold, test/functorch/test_vmap.py::TestVmapOperators::test_unsafe_view, test/functorch/test_vmap.py::TestVmapOperators::test_unsqueeze, test/functorch/test_vmap.py::TestVmapOperators::test_view, test/functorch/test_vmap.py::TestVmapOperators::test_view_as, test/functorch/test_vmap.py::TestVmapOperators::test_view_as_complex, test/functorch/test_vmap.py::TestVmapOperators::test_view_as_real, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_composition_in_dim_0_out_dim_0_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_composition_in_dim_0_out_dim_0_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_composition_in_dim_0_out_dim_1_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_composition_in_dim_0_out_dim_1_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_composition_in_dim_1_out_dim_0_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_composition_in_dim_1_out_dim_0_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_composition_in_dim_1_out_dim_1_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_composition_in_dim_1_out_dim_1_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_error_in_dim_0_out_dim_0_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_error_in_dim_0_out_dim_0_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_error_in_dim_0_out_dim_1_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_error_in_dim_0_out_dim_1_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_error_in_dim_1_out_dim_0_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_error_in_dim_1_out_dim_0_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_error_in_dim_1_out_dim_1_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_error_in_dim_1_out_dim_1_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_0_out_dim_0_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_0_out_dim_0_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_0_out_dim_1_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_0_out_dim_1_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_0_out_dim_2_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_0_out_dim_2_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_1_out_dim_0_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_1_out_dim_0_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_1_out_dim_1_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_1_out_dim_1_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_1_out_dim_2_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_1_out_dim_2_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_2_out_dim_0_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_2_out_dim_0_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_2_out_dim_1_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_2_out_dim_1_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_2_out_dim_2_randomness_error, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_chunksize_in_dim_2_out_dim_2_randomness_same, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_fallback_check, test/functorch/test_vmap.py::TestVmapOperators::test_vmap_fallback_check_ok, test/functorch/test_vmap.py::TestVmapOperators::test_weird_matmul_case, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_0d_tensor_index_put_inplace_False_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_0d_tensor_index_put_inplace_True_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_advanced_indexing_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_batch_norm_training_False_track_running_stats_False_affine_False_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_batch_norm_training_False_track_running_stats_False_affine_True_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_batch_norm_training_False_track_running_stats_True_affine_False_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_batch_norm_training_False_track_running_stats_True_affine_True_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_batch_norm_training_True_track_running_stats_False_affine_False_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_batch_norm_training_True_track_running_stats_False_affine_True_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_batch_norm_training_True_track_running_stats_True_affine_False_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_batch_norm_training_True_track_running_stats_True_affine_True_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_conv_double_backward_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_fill__Tensor_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_flatten_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_foo_like_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_group_norm_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_index_fill_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_index_put_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_inplace_on_view_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_isinf_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_isnan_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_linalg_eigh_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_linalg_svd_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_namedtuple_returns_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_nested_advanced_indexing_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_CubeGenVmapAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_ForwardHasDefaultArgsAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_H_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_MulGenVmapAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpyCatCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpyCubeAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpyCubeCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpyCubeNotComposableAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpyExpMarkDirtyAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpyMulAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpyMulCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpyMulScalarCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpyNMSCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpyNonzeroCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpySortAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpySortCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpySplitCopyCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpySplitCopyWithIntCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpyTakeAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpyTakeCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_NumpyViewCopyCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_ScaleGradGenVmapAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_SelectAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_SelectGenVmapAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_SortGenVmapAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_T_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_ZeroGradientsGenVmapAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule___getitem___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule___getitem___functorch_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule___radd___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule___rand___cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule___rdiv___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule___rmatmul___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule___rmod___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule___rmul___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule___ror___cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule___rpow___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule___rsub___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule___rxor___cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule__batch_norm_with_update_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule__chunk_cat_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule__native_batch_norm_legit_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule__segment_reduce_lengths_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule__segment_reduce_offsets_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule__softmax_backward_data_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule__unsafe_masked_index_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule__unsafe_masked_index_put_accumulate_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule__upsample_bilinear2d_aa_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_abs_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_acos_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_acosh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_add_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_addbmm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_addcdiv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_addcmul_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_addmm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_addmm_decomposed_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_addmv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_addr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_alias_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_all_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_allclose_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_amax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_amin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_aminmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_angle_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_any_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_arange_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_argmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_argmin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_argsort_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_argwhere_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_as_strided_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_as_strided_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_as_strided_partial_views_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_as_strided_scatter_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_asin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_asinh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_atan2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_atan_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_atanh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_atleast_1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_atleast_2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_atleast_3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_baddbmm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_bernoulli_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_bfloat16_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_bfloat16_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_bincount_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_bitwise_and_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_bitwise_left_shift_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_bitwise_not_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_bitwise_or_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_bitwise_right_shift_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_bitwise_xor_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_block_diag_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_bmm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_bool_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_bool_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_broadcast_shapes_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_broadcast_tensors_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_broadcast_to_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_bucketize_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_byte_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_byte_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cartesian_prod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cat_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cauchy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cdist_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cdouble_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_ceil_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cfloat_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_chalf_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_char_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_char_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cholesky_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cholesky_inverse_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cholesky_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_chunk_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_clamp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_clamp_max_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_clamp_min_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_clone_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_column_stack_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_combinations_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_complex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_conj_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_conj_physical_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_constant_pad_nd_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_contiguous_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_copysign_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_corrcoef_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cos_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cosh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_count_nonzero_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cov_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cross_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cummax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cummin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cumprod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cumsum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_cumulative_trapezoid_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_deg2rad_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_diag_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_diag_embed_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_diagflat_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_diagonal_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_diagonal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_diagonal_scatter_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_diff_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_digamma_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_dist_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_div_floor_rounding_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_div_no_rounding_mode_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_div_trunc_rounding_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_dot_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_double_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_double_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_dsplit_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_dstack_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_einsum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_empty_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_empty_like_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_empty_permuted_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_empty_strided_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_eq_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_equal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_erf_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_erfc_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_erfinv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_exp2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_exp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_expand_as_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_expand_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_expand_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_expm1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_exponential_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_eye_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_fft2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_fft_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_fftn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_fftshift_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_hfft2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_hfft_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_hfftn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_ifft2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_ifft_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_ifftn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_ifftshift_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_ihfft2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_ihfft_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_ihfftn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_irfft2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_irfft_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_irfftn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_rfft2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_rfft_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fft_rfftn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fill_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_flatten_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_flip_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fliplr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_flipud_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_float_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_float_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_float_power_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_floor_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_floor_divide_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fmin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_fmod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_frac_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_frexp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_full_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_full_like_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_gather_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_gcd_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_ge_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_geometric_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_geqrf_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_gradient_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_grid_sampler_2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_gt_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_half_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_half_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_heaviside_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_histc_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_hsplit_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_hstack_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_hypot_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_i0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_igamma_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_igammac_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_imag_cuda_complex64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_index_add_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_index_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_index_fill_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_index_put_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_index_put_functorch_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_index_reduce_amax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_index_reduce_amin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_index_reduce_mean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_index_reduce_prod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_index_select_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_inner_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_int_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_int_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_isclose_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_isfinite_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_isin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_isinf_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_isnan_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_isneginf_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_isposinf_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_isreal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_istft_cuda_complex64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_item_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_jiterator_2inputs_2outputs_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_jiterator_4inputs_with_extra_args_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_jiterator_binary_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_jiterator_binary_return_by_ref_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_jiterator_unary_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_kron_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_kthvalue_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_lcm_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_ldexp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_le_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_lerp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_lgamma_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_cholesky_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_cholesky_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_cond_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_cross_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_det_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_det_singular_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_diagonal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_eig_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_eigh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_eigvals_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_eigvalsh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_householder_product_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_inv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_inv_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_ldl_factor_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_ldl_factor_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_ldl_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_lstsq_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_lstsq_grad_oriented_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_lu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_lu_factor_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_lu_factor_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_lu_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_matrix_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_matrix_power_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_matrix_rank_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_matrix_rank_hermitian_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_multi_dot_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_norm_subgradients_at_zero_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_pinv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_pinv_hermitian_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_pinv_singular_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_qr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_slogdet_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_solve_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_solve_triangular_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_svd_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_svdvals_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_tensorinv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_tensorsolve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_vander_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_vecdot_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linalg_vector_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linspace_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_linspace_tensor_overload_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_log10_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_log1p_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_log2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_log_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_log_normal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_log_softmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_log_softmax_with_dtype_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_logaddexp2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_logaddexp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_logcumsumexp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_logdet_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_logical_and_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_logical_not_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_logical_or_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_logical_xor_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_logit_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_logspace_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_logspace_tensor_overload_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_logsumexp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_long_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_long_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_lt_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_lu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_lu_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_lu_unpack_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_mH_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_mT_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_amax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_amin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_argmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_argmin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_cumprod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_cumsum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_fill_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_fill_functorch_Scalar_only_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_log_softmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_logaddexp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_logsumexp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_mean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_median_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_normalize_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_prod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_scatter_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_select_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_softmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_softmin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_std_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_sum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_masked_var_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_matmul_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_matrix_exp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_max_binary_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_max_pool2d_with_indices_backward_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_max_reduction_no_dim_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_max_reduction_with_dim_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_maximum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_mean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_median_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_meshgrid_list_of_tensors_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_meshgrid_variadic_tensors_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_min_binary_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_min_reduction_no_dim_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_min_reduction_with_dim_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_minimum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_mm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_mode_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_movedim_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_msort_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_mul_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_multinomial_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_mv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_mvlgamma_mvlgamma_p_1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_mvlgamma_mvlgamma_p_3_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_mvlgamma_mvlgamma_p_5_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nan_to_num_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nanmean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nanmedian_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nanquantile_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nansum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_narrow_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_narrow_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_native_batch_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_native_dropout_backward_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_native_layer_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_ne_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_neg_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_new_empty_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_new_empty_strided_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_new_full_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_new_ones_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_new_zeros_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nextafter_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_adaptive_avg_pool1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_adaptive_avg_pool2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_adaptive_avg_pool3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_adaptive_max_pool1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_adaptive_max_pool2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_adaptive_max_pool3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_alpha_dropout_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_avg_pool1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_avg_pool2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_avg_pool3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_batch_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_batch_norm_without_cudnn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_bilinear_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_binary_cross_entropy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_binary_cross_entropy_with_logits_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_celu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_channel_shuffle_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv2d_no_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv2d_stride_depthwise_with_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv2d_stride_groups_with_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv2d_stride_no_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv2d_stride_padding_no_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv2d_stride_padding_with_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv2d_stride_with_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv2d_strided_padding_dilation_no_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv2d_strided_padding_dilation_with_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv2d_with_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv_transpose1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv_transpose2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_conv_transpose3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_cosine_embedding_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_cosine_similarity_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_cross_entropy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_ctc_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_dropout2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_dropout3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_dropout_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_elu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_embedding_bag_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_embedding_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_embedding_functorch_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_feature_alpha_dropout_with_train_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_feature_alpha_dropout_without_train_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_fractional_max_pool2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_fractional_max_pool3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_gaussian_nll_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_gelu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_glu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_grid_sample_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_group_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_hardshrink_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_hardsigmoid_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_hardswish_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_hardtanh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_hinge_embedding_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_huber_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_instance_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_interpolate_area_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_interpolate_bicubic_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_interpolate_bilinear_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_interpolate_linear_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_interpolate_nearest-exact_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_interpolate_nearest_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_interpolate_trilinear_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_kl_div_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_l1_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_layer_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_leaky_relu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_linear_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_local_response_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_logsigmoid_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_margin_ranking_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_max_pool1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_max_pool2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_max_pool3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_max_unpool1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_max_unpool1d_grad_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_max_unpool2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_max_unpool2d_grad_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_max_unpool3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_max_unpool3d_grad_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_mish_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_mse_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_mse_loss_functorch_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_multi_head_attention_forward_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_multi_margin_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_multilabel_margin_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_multilabel_soft_margin_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_nll_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_normalize_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_one_hot_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_pad_circular_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_pad_constant_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_pad_reflect_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_pad_replicate_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_pad_replicate_negative_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_pairwise_distance_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_pdist_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_pixel_shuffle_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_pixel_unshuffle_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_poisson_nll_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_prelu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_relu6_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_relu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_rms_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_rrelu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_scaled_dot_product_attention_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_selu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_silu_complex_cuda_complex64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_silu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_smooth_l1_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_soft_margin_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_softmin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_softmin_with_dtype_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_softplus_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_softshrink_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_softsign_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_tanhshrink_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_threshold_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_triplet_margin_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_triplet_margin_with_distance_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_unfold_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_upsample_bilinear_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nn_functional_upsample_nearest_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nonzero_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_nonzero_static_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_norm_fro_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_norm_inf_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_norm_nuc_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_normal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_normal_in_place_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_normal_number_mean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_ones_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_ones_like_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_ops_aten__new_zeros_with_same_feature_meta_functorchonly_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_ops_aten_index_put_functorch_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_ormqr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_outer_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_pca_lowrank_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_permute_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_pinverse_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_polar_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_polygamma_polygamma_n_0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_polygamma_polygamma_n_1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_polygamma_polygamma_n_2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_polygamma_polygamma_n_3_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_polygamma_polygamma_n_4_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_positive_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_pow_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_prod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_put_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_qr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_quantile_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_rad2deg_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_rand_like_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_randint_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_randint_like_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_randn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_randn_like_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_ravel_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_real_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_reciprocal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_remainder_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_renorm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_repeat_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_repeat_interleave_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_reshape_as_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_reshape_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_resize__cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_resize_as__cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_resolve_conj_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_resolve_neg_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_roll_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_rot90_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_round_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_round_decimals_0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_round_decimals_3_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_round_decimals_neg_3_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_rsqrt_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_rsub_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_scalar_tensor_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_scatter_add_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_scatter_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_scatter_reduce_amax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_scatter_reduce_amin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_scatter_reduce_mean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_scatter_reduce_prod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_scatter_reduce_sum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_searchsorted_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_select_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_select_scatter_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_sgn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_short_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_short_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_sigmoid_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_sign_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_signal_windows_bartlett_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_signal_windows_blackman_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_signal_windows_cosine_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_signal_windows_exponential_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_signal_windows_gaussian_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_signal_windows_general_cosine_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_signal_windows_general_hamming_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_signal_windows_hamming_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_signal_windows_hann_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_signal_windows_kaiser_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_signal_windows_nuttall_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_signbit_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_sin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_sinc_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_sinh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_slice_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_slice_scatter_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_softmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_softmax_with_dtype_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_sort_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_sparse_mm_reduce_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_sparse_sampled_addmm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_airy_ai_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_bessel_j0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_bessel_j1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_bessel_y0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_bessel_y1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_chebyshev_polynomial_t_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_chebyshev_polynomial_u_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_chebyshev_polynomial_v_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_chebyshev_polynomial_w_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_entr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_erfcx_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_hermite_polynomial_h_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_hermite_polynomial_he_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_i0e_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_i1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_i1e_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_laguerre_polynomial_l_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_legendre_polynomial_p_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_log_ndtr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_modified_bessel_i0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_modified_bessel_i1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_modified_bessel_k0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_modified_bessel_k1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_ndtr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_ndtri_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_polygamma_special_polygamma_n_0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_scaled_modified_bessel_k0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_scaled_modified_bessel_k1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_shifted_chebyshev_polynomial_t_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_shifted_chebyshev_polynomial_u_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_shifted_chebyshev_polynomial_v_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_shifted_chebyshev_polynomial_w_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_spherical_bessel_j0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_xlog1py_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_special_zeta_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_split_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_split_list_args_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_split_with_sizes_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_split_with_sizes_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_sqrt_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_square_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_squeeze_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_squeeze_multiple_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_stack_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_std_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_std_mean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_std_mean_unbiased_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_std_unbiased_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_stft_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_sub_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_sum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_sum_to_size_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_svd_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_svd_lowrank_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_t_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_t_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_take_along_dim_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_take_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_tan_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_tanh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_tensor_split_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_tensordot_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_tile_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_to_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_to_sparse_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_topk_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_torch_ops_aten__efficient_attention_forward_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_torch_ops_aten__flash_attention_forward_cuda_float16, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_torch_ops_aten__safe_softmax_default_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_trace_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_transpose_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_trapezoid_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_trapz_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_triangular_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_tril_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_tril_indices_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_triu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_triu_indices_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_true_divide_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_trunc_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_unbind_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_unflatten_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_unfold_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_unfold_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_uniform_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_unique_consecutive_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_unique_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_unravel_index_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_unsafe_chunk_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_unsafe_split_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_unsqueeze_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_unsqueeze_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_var_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_var_mean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_var_mean_unbiased_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_var_unbiased_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_vdot_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_view_as_complex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_view_as_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_view_as_real_cuda_complex64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_view_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_view_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_vsplit_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_vstack_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_where_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_xlogy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_zero__cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_zeros_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_op_has_batch_rule_zeros_like_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_searchsorted_bucketize_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_slogdet_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_sum_scalar_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_torch_return_types_returns_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_escaped_error_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_CubeGenVmapAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_ForwardHasDefaultArgsAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_H_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_MulGenVmapAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpyCatCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpyCubeAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpyCubeCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpyCubeNotComposableAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpyExpMarkDirtyAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpyMulAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpyMulCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpyMulScalarCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpyNMSCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpyNonzeroCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpySortAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpySortCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpySplitCopyCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpySplitCopyWithIntCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpyTakeAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpyTakeCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_NumpyViewCopyCustomOp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_ScaleGradGenVmapAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_SelectAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_SelectGenVmapAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_SortGenVmapAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_T_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_ZeroGradientsGenVmapAutogradFunction_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive___getitem___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive___getitem___functorch_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive___radd___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive___rand___cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive___rdiv___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive___rmatmul___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive___rmod___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive___rmul___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive___ror___cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive___rpow___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive___rsub___cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive___rxor___cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive__batch_norm_with_update_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive__chunk_cat_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive__native_batch_norm_legit_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive__segment_reduce_lengths_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive__segment_reduce_offsets_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive__softmax_backward_data_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive__unsafe_masked_index_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive__unsafe_masked_index_put_accumulate_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive__upsample_bilinear2d_aa_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_abs_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_acos_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_acosh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_add_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_addbmm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_addcdiv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_addcmul_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_addmm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_addmm_decomposed_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_addmv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_addr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_alias_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_all_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_allclose_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_amax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_amin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_aminmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_angle_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_any_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_arange_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_argmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_argmin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_argsort_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_argwhere_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_as_strided_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_as_strided_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_as_strided_partial_views_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_as_strided_scatter_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_asin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_asinh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_atan2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_atan_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_atanh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_atleast_1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_atleast_2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_atleast_3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_baddbmm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_bernoulli_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_bfloat16_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_bfloat16_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_bincount_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_bitwise_and_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_bitwise_left_shift_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_bitwise_not_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_bitwise_or_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_bitwise_right_shift_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_bitwise_xor_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_block_diag_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_bmm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_bool_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_bool_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_broadcast_shapes_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_broadcast_tensors_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_broadcast_to_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_bucketize_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_byte_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_byte_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cartesian_prod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cat_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cauchy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cdist_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cdouble_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_ceil_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cfloat_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_chalf_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_char_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_char_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cholesky_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cholesky_inverse_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cholesky_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_chunk_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_clamp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_clamp_max_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_clamp_min_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_clone_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_column_stack_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_combinations_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_complex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_conj_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_conj_physical_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_constant_pad_nd_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_contiguous_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_copysign_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_corrcoef_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cos_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cosh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_count_nonzero_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cov_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cross_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cummax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cummin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cumprod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cumsum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_cumulative_trapezoid_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_deg2rad_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_diag_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_diag_embed_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_diagflat_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_diagonal_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_diagonal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_diagonal_scatter_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_diff_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_digamma_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_dist_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_div_floor_rounding_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_div_no_rounding_mode_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_div_trunc_rounding_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_dot_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_double_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_double_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_dsplit_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_dstack_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_einsum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_empty_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_empty_like_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_empty_permuted_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_empty_strided_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_eq_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_equal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_erf_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_erfc_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_erfinv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_exp2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_exp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_expand_as_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_expand_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_expand_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_expm1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_exponential_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_eye_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_fft2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_fft_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_fftn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_fftshift_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_hfft2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_hfft_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_hfftn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_ifft2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_ifft_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_ifftn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_ifftshift_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_ihfft2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_ihfft_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_ihfftn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_irfft2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_irfft_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_irfftn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_rfft2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_rfft_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fft_rfftn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fill_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_flatten_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_flip_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fliplr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_flipud_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_float_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_float_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_float_power_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_floor_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_floor_divide_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fmin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_fmod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_frac_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_frexp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_full_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_full_like_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_gather_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_gcd_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_ge_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_geometric_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_geqrf_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_gradient_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_grid_sampler_2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_gt_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_half_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_half_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_heaviside_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_histc_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_hsplit_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_hstack_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_hypot_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_i0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_igamma_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_igammac_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_imag_cuda_complex64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_index_add_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_index_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_index_fill_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_index_put_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_index_put_functorch_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_index_reduce_amax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_index_reduce_amin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_index_reduce_mean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_index_reduce_prod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_index_select_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_inner_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_int_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_int_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_isclose_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_isfinite_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_isin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_isinf_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_isnan_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_isneginf_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_isposinf_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_isreal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_istft_cuda_complex64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_item_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_jiterator_2inputs_2outputs_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_jiterator_4inputs_with_extra_args_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_jiterator_binary_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_jiterator_binary_return_by_ref_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_jiterator_unary_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_kron_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_kthvalue_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_lcm_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_ldexp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_le_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_lerp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_lgamma_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_cholesky_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_cholesky_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_cond_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_cross_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_det_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_det_singular_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_diagonal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_eig_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_eigh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_eigvals_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_eigvalsh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_householder_product_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_inv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_inv_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_ldl_factor_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_ldl_factor_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_ldl_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_lstsq_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_lstsq_grad_oriented_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_lu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_lu_factor_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_lu_factor_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_lu_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_matrix_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_matrix_power_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_matrix_rank_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_matrix_rank_hermitian_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_multi_dot_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_norm_subgradients_at_zero_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_pinv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_pinv_hermitian_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_pinv_singular_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_qr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_slogdet_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_solve_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_solve_triangular_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_svd_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_svdvals_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_tensorinv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_tensorsolve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_vander_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_vecdot_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linalg_vector_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linspace_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_linspace_tensor_overload_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_log10_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_log1p_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_log2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_log_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_log_normal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_log_softmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_log_softmax_with_dtype_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_logaddexp2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_logaddexp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_logcumsumexp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_logdet_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_logical_and_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_logical_not_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_logical_or_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_logical_xor_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_logit_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_logspace_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_logspace_tensor_overload_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_logsumexp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_long_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_long_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_lt_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_lu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_lu_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_lu_unpack_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_mH_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_mT_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_amax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_amin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_argmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_argmin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_cumprod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_cumsum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_fill_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_fill_functorch_Scalar_only_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_log_softmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_logaddexp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_logsumexp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_mean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_median_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_normalize_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_prod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_scatter_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_select_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_softmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_softmin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_std_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_sum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_masked_var_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_matmul_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_matrix_exp_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_max_binary_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_max_pool2d_with_indices_backward_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_max_reduction_no_dim_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_max_reduction_with_dim_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_maximum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_mean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_median_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_meshgrid_list_of_tensors_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_meshgrid_variadic_tensors_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_min_binary_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_min_reduction_no_dim_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_min_reduction_with_dim_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_minimum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_mm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_mode_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_movedim_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_msort_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_mul_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_multinomial_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_mv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_mvlgamma_mvlgamma_p_1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_mvlgamma_mvlgamma_p_3_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_mvlgamma_mvlgamma_p_5_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nan_to_num_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nanmean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nanmedian_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nanquantile_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nansum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_narrow_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_narrow_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_native_batch_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_native_dropout_backward_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_native_layer_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_ne_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_neg_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_new_empty_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_new_empty_strided_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_new_full_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_new_ones_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_new_zeros_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nextafter_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_adaptive_avg_pool1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_adaptive_avg_pool2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_adaptive_avg_pool3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_adaptive_max_pool1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_adaptive_max_pool2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_adaptive_max_pool3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_alpha_dropout_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_avg_pool1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_avg_pool2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_avg_pool3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_batch_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_batch_norm_without_cudnn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_bilinear_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_binary_cross_entropy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_binary_cross_entropy_with_logits_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_celu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_channel_shuffle_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv2d_no_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv2d_stride_depthwise_with_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv2d_stride_groups_with_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv2d_stride_no_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv2d_stride_padding_no_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv2d_stride_padding_with_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv2d_stride_with_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv2d_strided_padding_dilation_no_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv2d_strided_padding_dilation_with_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv2d_with_bias_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv_transpose1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv_transpose2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_conv_transpose3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_cosine_embedding_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_cosine_similarity_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_cross_entropy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_ctc_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_dropout2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_dropout3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_dropout_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_elu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_embedding_bag_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_embedding_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_embedding_functorch_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_feature_alpha_dropout_with_train_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_feature_alpha_dropout_without_train_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_fractional_max_pool2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_fractional_max_pool3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_gaussian_nll_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_gelu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_glu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_grid_sample_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_group_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_hardshrink_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_hardsigmoid_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_hardswish_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_hardtanh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_hinge_embedding_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_huber_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_instance_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_interpolate_area_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_interpolate_bicubic_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_interpolate_bilinear_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_interpolate_linear_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_interpolate_nearest-exact_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_interpolate_nearest_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_interpolate_trilinear_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_kl_div_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_l1_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_layer_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_leaky_relu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_linear_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_local_response_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_logsigmoid_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_margin_ranking_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_max_pool1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_max_pool2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_max_pool3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_max_unpool1d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_max_unpool1d_grad_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_max_unpool2d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_max_unpool2d_grad_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_max_unpool3d_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_max_unpool3d_grad_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_mish_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_mse_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_mse_loss_functorch_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_multi_head_attention_forward_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_multi_margin_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_multilabel_margin_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_multilabel_soft_margin_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_nll_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_normalize_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_one_hot_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_pad_circular_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_pad_constant_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_pad_reflect_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_pad_replicate_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_pad_replicate_negative_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_pairwise_distance_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_pdist_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_pixel_shuffle_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_pixel_unshuffle_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_poisson_nll_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_prelu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_relu6_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_relu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_rms_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_rrelu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_scaled_dot_product_attention_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_selu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_silu_complex_cuda_complex64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_silu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_smooth_l1_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_soft_margin_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_softmin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_softmin_with_dtype_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_softplus_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_softshrink_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_softsign_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_tanhshrink_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_threshold_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_triplet_margin_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_triplet_margin_with_distance_loss_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_unfold_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_upsample_bilinear_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nn_functional_upsample_nearest_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nonzero_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_nonzero_static_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_norm_fro_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_norm_inf_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_norm_nuc_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_normal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_normal_in_place_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_normal_number_mean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_ones_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_ones_like_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_ops_aten__new_zeros_with_same_feature_meta_functorchonly_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_ops_aten_index_put_functorch_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_ormqr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_outer_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_pca_lowrank_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_permute_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_pinverse_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_polar_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_polygamma_polygamma_n_0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_polygamma_polygamma_n_1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_polygamma_polygamma_n_2_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_polygamma_polygamma_n_3_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_polygamma_polygamma_n_4_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_positive_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_pow_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_prod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_put_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_qr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_quantile_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_rad2deg_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_rand_like_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_randint_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_randint_like_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_randn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_randn_like_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_ravel_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_real_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_reciprocal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_remainder_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_renorm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_repeat_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_repeat_interleave_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_reshape_as_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_reshape_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_resize__cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_resize_as__cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_resolve_conj_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_resolve_neg_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_roll_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_rot90_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_round_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_round_decimals_0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_round_decimals_3_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_round_decimals_neg_3_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_rsqrt_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_rsub_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_scalar_tensor_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_scatter_add_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_scatter_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_scatter_reduce_amax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_scatter_reduce_amin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_scatter_reduce_mean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_scatter_reduce_prod_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_scatter_reduce_sum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_searchsorted_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_select_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_select_scatter_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_sgn_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_short_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_short_functorch_no_channels_last_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_sigmoid_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_sign_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_signal_windows_bartlett_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_signal_windows_blackman_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_signal_windows_cosine_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_signal_windows_exponential_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_signal_windows_gaussian_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_signal_windows_general_cosine_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_signal_windows_general_hamming_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_signal_windows_hamming_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_signal_windows_hann_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_signal_windows_kaiser_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_signal_windows_nuttall_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_signbit_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_sin_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_sinc_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_sinh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_slice_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_slice_scatter_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_softmax_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_softmax_with_dtype_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_sort_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_sparse_mm_reduce_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_sparse_sampled_addmm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_airy_ai_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_bessel_j0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_bessel_j1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_bessel_y0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_bessel_y1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_chebyshev_polynomial_t_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_chebyshev_polynomial_u_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_chebyshev_polynomial_v_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_chebyshev_polynomial_w_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_entr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_erfcx_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_hermite_polynomial_h_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_hermite_polynomial_he_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_i0e_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_i1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_i1e_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_laguerre_polynomial_l_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_legendre_polynomial_p_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_log_ndtr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_modified_bessel_i0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_modified_bessel_i1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_modified_bessel_k0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_modified_bessel_k1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_ndtr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_ndtri_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_polygamma_special_polygamma_n_0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_scaled_modified_bessel_k0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_scaled_modified_bessel_k1_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_shifted_chebyshev_polynomial_t_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_shifted_chebyshev_polynomial_u_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_shifted_chebyshev_polynomial_v_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_shifted_chebyshev_polynomial_w_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_spherical_bessel_j0_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_xlog1py_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_special_zeta_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_split_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_split_list_args_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_split_with_sizes_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_split_with_sizes_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_sqrt_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_square_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_squeeze_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_squeeze_multiple_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_stack_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_std_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_std_mean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_std_mean_unbiased_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_std_unbiased_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_stft_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_sub_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_sum_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_sum_to_size_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_svd_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_svd_lowrank_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_t_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_t_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_take_along_dim_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_take_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_tan_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_tanh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_tensor_split_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_tensordot_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_tile_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_to_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_to_sparse_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_topk_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_torch_ops_aten__efficient_attention_forward_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_torch_ops_aten__flash_attention_forward_cuda_float16, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_torch_ops_aten__safe_softmax_default_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_trace_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_transpose_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_trapezoid_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_trapz_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_triangular_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_tril_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_tril_indices_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_triu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_triu_indices_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_true_divide_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_trunc_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_unbind_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_unflatten_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_unfold_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_unfold_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_uniform_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_unique_consecutive_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_unique_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_unravel_index_cuda_int64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_unsafe_chunk_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_unsafe_split_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_unsqueeze_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_unsqueeze_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_var_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_var_mean_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_var_mean_unbiased_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_var_unbiased_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_vdot_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_view_as_complex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_view_as_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_view_as_real_cuda_complex64, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_view_copy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_view_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_vsplit_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_vstack_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_where_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_xlogy_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_zero__cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_zeros_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_exhaustive_zeros_like_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_cholesky_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_cholesky_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_cond_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_cross_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_det_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_det_singular_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_diagonal_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_eig_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_eigh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_eigvals_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_eigvalsh_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_householder_product_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_inv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_inv_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_ldl_factor_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_ldl_factor_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_ldl_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_lstsq_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_lstsq_grad_oriented_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_lu_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_lu_factor_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_lu_factor_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_lu_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_matrix_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_matrix_power_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_matrix_rank_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_matrix_rank_hermitian_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_multi_dot_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_norm_subgradients_at_zero_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_pinv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_pinv_hermitian_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_pinv_singular_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_qr_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_slogdet_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_solve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_solve_ex_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_solve_triangular_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_svd_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_svdvals_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_tensorinv_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_tensorsolve_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_vander_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_vecdot_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_linalg_failure_1D_input_linalg_vector_norm_cuda_float32, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_multi_dot_failure_1D_input_cuda, test/functorch/test_vmap.py::TestVmapOperatorsOpInfoCUDA::test_vmap_with_anomaly_detection_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_add_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_binary_cross_entropy_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_diagonal_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_div_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_expand_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_index_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_inplace_manyview_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_inplace_view_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_lgamma_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_log1p_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_log_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_log_softmax_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_logsumexp_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_max_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_median_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_min_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_mul_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_permute_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_reshape_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_select_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_sigmoid_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_slice_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_stack_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_sub_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_threshold_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_trace_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_unrelated_output_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_unrelated_output_multiple_grad_cuda, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_vmap_fallback_check, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_vmap_fallback_check_ok, test/functorch/test_vmap.py::TestVmapBatchedGradientCUDA::test_where_cuda, test/functorch/test_vmap.py::TestTransformFailureCUDA::test_fails_with_autograd_function_transform_grad_and_value_cuda, test/functorch/test_vmap.py::TestTransformFailureCUDA::test_fails_with_autograd_function_transform_grad_cuda, test/functorch/test_vmap.py::TestTransformFailureCUDA::test_fails_with_autograd_function_transform_jacfwd_cuda, test/functorch/test_vmap.py::TestTransformFailureCUDA::test_fails_with_autograd_function_transform_jacrev_cuda, test/functorch/test_vmap.py::TestTransformFailureCUDA::test_fails_with_autograd_function_transform_jvp_cuda, test/functorch/test_vmap.py::TestTransformFailureCUDA::test_fails_with_autograd_function_transform_vjp_cuda, test/functorch/test_vmap.py::TestTransformFailureCUDA::test_fails_with_autograd_function_transform_vmap_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_alpha_dropout_randomness_different_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_alpha_dropout_randomness_different_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_alpha_dropout_randomness_different_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_alpha_dropout_randomness_error_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_alpha_dropout_randomness_error_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_alpha_dropout_randomness_error_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_alpha_dropout_randomness_same_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_alpha_dropout_randomness_same_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_alpha_dropout_randomness_same_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_different_batched_input_first_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_different_batched_input_first_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_different_batched_input_first_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_different_batched_input_last_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_different_batched_input_last_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_different_batched_input_last_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_different_batched_input_none_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_different_batched_input_none_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_different_batched_input_none_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_error_batched_input_first_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_error_batched_input_first_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_error_batched_input_first_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_error_batched_input_last_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_error_batched_input_last_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_error_batched_input_last_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_error_batched_input_none_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_error_batched_input_none_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_error_batched_input_none_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_same_batched_input_first_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_same_batched_input_first_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_same_batched_input_first_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_same_batched_input_last_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_same_batched_input_last_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_same_batched_input_last_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_same_batched_input_none_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_same_batched_input_none_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_False_randomness_same_batched_input_none_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_different_batched_input_first_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_different_batched_input_first_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_different_batched_input_first_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_different_batched_input_last_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_different_batched_input_last_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_different_batched_input_last_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_different_batched_input_none_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_different_batched_input_none_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_different_batched_input_none_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_error_batched_input_first_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_error_batched_input_first_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_error_batched_input_first_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_error_batched_input_last_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_error_batched_input_last_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_error_batched_input_last_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_error_batched_input_none_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_error_batched_input_none_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_error_batched_input_none_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_same_batched_input_first_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_same_batched_input_first_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_same_batched_input_first_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_same_batched_input_last_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_same_batched_input_last_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_same_batched_input_last_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_same_batched_input_none_batched_probability_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_same_batched_input_none_batched_probability_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_bernoulli_in_place_use_generator_True_randomness_same_batched_input_none_batched_probability_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_chunk_vmap_in_dim_0_out_dim_0_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_chunk_vmap_in_dim_0_out_dim_1_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_chunk_vmap_in_dim_0_out_dim_2_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_chunk_vmap_in_dim_1_out_dim_0_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_chunk_vmap_in_dim_1_out_dim_1_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_chunk_vmap_in_dim_1_out_dim_2_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_chunk_vmap_in_dim_2_out_dim_0_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_chunk_vmap_in_dim_2_out_dim_1_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_chunk_vmap_in_dim_2_out_dim_2_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_dropout_randomness_different_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_dropout_randomness_different_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_dropout_randomness_different_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_dropout_randomness_error_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_dropout_randomness_error_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_dropout_randomness_error_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_dropout_randomness_same_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_dropout_randomness_same_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_dropout_randomness_same_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_dropout_unbatched_randomness_different_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_dropout_unbatched_randomness_error_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_dropout_unbatched_randomness_same_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_factory_ops_randomness_different_use_generator_False_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_factory_ops_randomness_different_use_generator_True_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_factory_ops_randomness_error_use_generator_False_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_factory_ops_randomness_error_use_generator_True_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_factory_ops_randomness_same_use_generator_False_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_factory_ops_randomness_same_use_generator_True_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_alpha_dropout_randomness_different_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_alpha_dropout_randomness_different_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_alpha_dropout_randomness_different_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_alpha_dropout_randomness_error_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_alpha_dropout_randomness_error_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_alpha_dropout_randomness_error_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_alpha_dropout_randomness_same_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_alpha_dropout_randomness_same_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_alpha_dropout_randomness_same_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_different_batched_input_first_dim_2_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_different_batched_input_first_dim_3_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_different_batched_input_last_dim_2_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_different_batched_input_last_dim_3_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_different_batched_input_none_dim_2_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_different_batched_input_none_dim_3_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_error_batched_input_first_dim_2_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_error_batched_input_first_dim_3_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_error_batched_input_last_dim_2_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_error_batched_input_last_dim_3_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_error_batched_input_none_dim_2_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_error_batched_input_none_dim_3_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_same_batched_input_first_dim_2_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_same_batched_input_first_dim_3_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_same_batched_input_last_dim_2_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_same_batched_input_last_dim_3_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_same_batched_input_none_dim_2_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_feature_dropout_randomness_same_batched_input_none_dim_3_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_jacfwd_with_random_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_like_functions_randomness_different_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_like_functions_randomness_different_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_like_functions_randomness_different_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_like_functions_randomness_error_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_like_functions_randomness_error_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_like_functions_randomness_error_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_like_functions_randomness_same_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_like_functions_randomness_same_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_like_functions_randomness_same_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_different_batched_call_False_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_different_batched_call_False_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_different_batched_call_False_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_different_batched_call_True_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_different_batched_call_True_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_different_batched_call_True_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_error_batched_call_False_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_error_batched_call_False_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_error_batched_call_False_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_error_batched_call_True_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_error_batched_call_True_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_error_batched_call_True_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_same_batched_call_False_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_same_batched_call_False_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_same_batched_call_False_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_same_batched_call_True_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_same_batched_call_True_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_False_randomness_same_batched_call_True_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_different_batched_call_False_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_different_batched_call_False_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_different_batched_call_False_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_different_batched_call_True_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_different_batched_call_True_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_different_batched_call_True_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_error_batched_call_False_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_error_batched_call_False_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_error_batched_call_False_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_error_batched_call_True_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_error_batched_call_True_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_error_batched_call_True_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_same_batched_call_False_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_same_batched_call_False_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_same_batched_call_False_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_same_batched_call_True_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_same_batched_call_True_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_multinomial_use_generator_True_randomness_same_batched_call_True_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_different_batched_input_first_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_different_batched_input_first_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_different_batched_input_first_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_different_batched_input_last_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_different_batched_input_last_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_different_batched_input_last_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_different_batched_input_none_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_different_batched_input_none_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_different_batched_input_none_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_error_batched_input_first_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_error_batched_input_first_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_error_batched_input_first_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_error_batched_input_last_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_error_batched_input_last_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_error_batched_input_last_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_error_batched_input_none_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_error_batched_input_none_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_error_batched_input_none_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_same_batched_input_first_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_same_batched_input_first_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_same_batched_input_first_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_same_batched_input_last_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_same_batched_input_last_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_same_batched_input_last_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_same_batched_input_none_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_same_batched_input_none_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_False_randomness_same_batched_input_none_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_different_batched_input_first_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_different_batched_input_first_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_different_batched_input_first_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_different_batched_input_last_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_different_batched_input_last_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_different_batched_input_last_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_different_batched_input_none_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_different_batched_input_none_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_different_batched_input_none_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_error_batched_input_first_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_error_batched_input_first_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_error_batched_input_first_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_error_batched_input_last_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_error_batched_input_last_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_error_batched_input_last_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_error_batched_input_none_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_error_batched_input_none_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_error_batched_input_none_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_same_batched_input_first_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_same_batched_input_first_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_same_batched_input_first_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_same_batched_input_last_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_same_batched_input_last_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_same_batched_input_last_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_same_batched_input_none_batched_other_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_same_batched_input_none_batched_other_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_binary_out_of_place_use_generator_True_randomness_same_batched_input_none_batched_other_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_False_randomness_different_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_False_randomness_different_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_False_randomness_different_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_False_randomness_error_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_False_randomness_error_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_False_randomness_error_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_False_randomness_same_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_False_randomness_same_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_False_randomness_same_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_True_randomness_different_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_True_randomness_different_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_True_randomness_different_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_True_randomness_error_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_True_randomness_error_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_True_randomness_error_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_True_randomness_same_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_True_randomness_same_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_inplace_use_generator_True_randomness_same_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_False_randomness_different_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_False_randomness_different_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_False_randomness_different_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_False_randomness_error_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_False_randomness_error_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_False_randomness_error_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_False_randomness_same_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_False_randomness_same_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_False_randomness_same_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_True_randomness_different_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_True_randomness_different_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_True_randomness_different_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_True_randomness_error_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_True_randomness_error_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_True_randomness_error_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_True_randomness_same_batched_input_first_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_True_randomness_same_batched_input_last_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_random_unary_out_of_place_use_generator_True_randomness_same_batched_input_none_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_randperm_randomness_different_use_generator_False_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_randperm_randomness_different_use_generator_True_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_randperm_randomness_error_use_generator_False_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_randperm_randomness_error_use_generator_True_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_randperm_randomness_same_use_generator_False_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_randperm_randomness_same_use_generator_True_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_unsupported_random_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_vmap_chunksize_in_dim_0_out_dim_0_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_vmap_chunksize_in_dim_0_out_dim_1_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_vmap_chunksize_in_dim_0_out_dim_2_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_vmap_chunksize_in_dim_1_out_dim_0_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_vmap_chunksize_in_dim_1_out_dim_1_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_vmap_chunksize_in_dim_1_out_dim_2_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_vmap_chunksize_in_dim_2_out_dim_0_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_vmap_chunksize_in_dim_2_out_dim_1_cuda, test/functorch/test_vmap.py::TestRandomnessCUDA::test_vmap_chunksize_in_dim_2_out_dim_2_cuda, test/functorch/test_vmap.py::TestVmapDeviceTypeCUDA::test__is_all_true_cuda, test/functorch/test_vmap.py::TestVmapDeviceTypeCUDA::test__is_any_true_cuda, test/functorch/test_vmap.py::TestVmapDeviceTypeCUDA::test_check_tensor_cuda, test/functorch/test_vmap.py::TestVmapDeviceTypeCUDA::test_vmap_fallback_check, test/functorch/test_vmap.py::TestVmapDeviceTypeCUDA::test_vmap_fallback_check_ok, test/functorch/test_vmap.py::TestVmapNestedTensorCUDA::test_cat_batching_rule_cuda, test/functorch/test_vmap.py::TestVmapNestedTensorCUDA::test_fallback_binary_cuda, test/functorch/test_vmap.py::TestVmapNestedTensorCUDA::test_fallback_binary_nt_and_batched_dense_cuda, test/functorch/test_vmap.py::TestVmapNestedTensorCUDA::test_fallback_binary_nt_and_unbatched_dense_cuda, test/functorch/test_vmap.py::TestVmapNestedTensorCUDA::test_fallback_unary_cuda, test/functorch/test_vmap.py::TestVmapNestedTensorCUDA::test_fallback_with_nt_and_batched_dense_with_nonzero_bdim_raises_cuda, test/functorch/test_vmap.py::TestVmapNestedTensorCUDA::test_multilevel_vmap_raises_cuda, test/functorch/test_vmap.py::TestVmapNestedTensorCUDA::test_nt_acts_as_dense_in_vmap_cuda, test/functorch/test_vmap.py::TestVmapNestedTensorCUDA::test_nt_with_nonzero_in_dim_raises_cuda, test/functorch/test_vmap.py::TestVmapNestedTensorCUDA::test_nt_with_nonzero_out_dim_raises_cuda, test/functorch/test_vmap.py::TestVmapNestedTensorCUDA::test_shape_call_cuda, test/functorch/test_vmap.py::TestVmapNestedTensorCUDA::test_vmap_fallback_check, test/functorch/test_vmap.py::TestVmapNestedTensorCUDA::test_vmap_fallback_check_ok 2024-08-06T21:51:06.9293853Z 2024-08-06T21:51:10.4761371Z Running torch_np/test_binary_ufuncs 1/1 ... [2024-08-06 21:51:10.475579] 2024-08-06T21:51:10.4765315Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'torch_np/test_binary_ufuncs.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:51:10.476056] 2024-08-06T21:51:14.8991568Z 2024-08-06T21:51:14.8992678Z torch_np/test_binary_ufuncs 1/1 was successful, full logs can be found in artifacts with path test/test-reports/torch_np.test_binary_ufuncs_1.1_6f8c0ad317152332_.log 2024-08-06T21:51:14.9007504Z Running 38 items in this shard: test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_add, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_arctan2, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_bitwise_and, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_bitwise_or, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_bitwise_xor, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_copysign, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_divide, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_equal, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_float_power, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_floor_divide, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_fmax, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_fmin, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_fmod, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_gcd, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_greater, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_greater_equal, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_heaviside, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_hypot, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_lcm, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_ldexp, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_left_shift, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_less, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_less_equal, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_logaddexp, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_logaddexp2, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_logical_and, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_logical_or, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_logical_xor, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_matmul, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_maximum, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_minimum, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_multiply, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_nextafter, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_not_equal, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_power, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_remainder, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_right_shift, test/torch_np/test_binary_ufuncs.py::TestBinaryUfuncBasic::test_subtract 2024-08-06T21:51:14.9021058Z 2024-08-06T21:51:18.6010208Z Running dynamo/test_inline_inbuilt_nn_modules 1/1 ... [2024-08-06 21:51:18.600518] 2024-08-06T21:51:18.6015395Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'dynamo/test_inline_inbuilt_nn_modules.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:51:18.601056] 2024-08-06T21:54:20.5540620Z 2024-08-06T21:54:20.5542006Z dynamo/test_inline_inbuilt_nn_modules 1/1 was successful, full logs can be found in artifacts with path test/test-reports/dynamo.test_inline_inbuilt_nn_modules_1.1_3b1c8a8197223013_.log 2024-08-06T21:54:20.6309724Z Running 1102 items in this shard: test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_312_binary_slice_with_graph_break1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_312_binary_slice_with_graph_break2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_T_tensor_attribute_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_add_sizes_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_add_to_set_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_anomaly_aot_autograd_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_any_all_symnode_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_aot_autograd_propagate_unbacked_symints_shape_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_assert_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_assert_size_stride_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_assigning_function_to_class_attribute_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_assigning_function_to_object_attribute_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_auto_functionalize_can_with_default_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_auto_functionalize_can_with_none_return_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_auto_functionalize_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_auto_functionalize_on_view_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_auto_functionalize_optional_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_auto_functionalize_self_as_mutate_arg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_auto_functionalize_tensorlist_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_auto_functionalize_with_returns_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_backend_match_guard_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_backend_match_guard_multi_threads_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_backward_deterministic_mode_mismatch_warning_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_boolarg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_build_tuple_unpack_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_builder_for_class_with_metaclass_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_builtin_abs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_builtin_isinstance_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_builtin_str_on_user_defined_function_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_builtin_subclasses_as_method_on_class_type_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_builtin_subclasses_as_method_on_var_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_call_parent_non_class_methods_from_child_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_callpacked_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_can_auto_functionalize_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cannot_trace_mark_dynamic_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cannot_trace_mark_dynamic_safe_unreached_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cast_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cat_unbacked_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_catch_watchings1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_catch_watchings2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cell_output1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cell_output2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_class_duner_mro_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_class_has_instancecheck_method_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_clone_sparse_input_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_closure_out_of_scope_cell_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_closure_out_of_scope_cell_with_cond_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_closure_out_of_scope_cell_with_mutation_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_closure_recompiles_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_closure_with_mutation_and_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_compare_shapes_eq_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_compare_shapes_neq_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_compare_shapes_tuple_eq_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_compare_shapes_tuple_neq_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_compare_shapes_with_constant_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_compilation_metrics_size_limit_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_compile_profiler_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cond_export_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cond_export_single_arg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cond_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cond_nested_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cond_side_effects_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cond_with_quantization_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_conditional_list_comp_in_context_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_config_getattr_default_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_config_obj_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_const_dict_variable_python_type_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_constant_getattr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_contains_dunder_dict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cpp_extension_recommends_custom_ops_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cross_entropy_loss_fancy_ctor1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cross_entropy_loss_fancy_ctor2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cross_entropy_loss_simple_ctor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cse_dict_guards_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_cuda_set_device_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_custom_dict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_custom_iter_dict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_custom_keys_iter_dict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_custom_module_free_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dataclass_fields_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dataclass_local_hasattr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_default_args_device_dtype_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_default_dtype_change_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_defaultdict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_deque_append_left_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_deque_input_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_derpy_nn_module_usage_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_descriptor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_deterministic_algorithms_mutated_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dict_guard_on_keys_order2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dict_guard_on_keys_order_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dict_mutation_side_effect_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dict_namedtuple_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dict_order_keys_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dict_order_keys_modules_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dict_order_keys_tensors_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dict_reconstruct_keeps_original_order_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dict_subclass_cannot_be_initialized_in_graph_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dictcomp_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_disable_flag_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dtypes_no_graphbreaks_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dunder_methods_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dunder_new_function_inlining_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_duplicate_graph_break_log_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dynamic_one_hot_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dynamo_cache_invalidate_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dynamo_cache_move_to_front_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dynamo_compiling_fake_tensor_to_vararg_int_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dynamo_min_operator_with_shape_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_dynamo_reset_clears_cache_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_empty_list_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_enum_as_dict_key_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_enum_as_dict_key_with_overloaded_str_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_enum_guards_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_enum_no_graphbreaks_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_error_on_nested_fx_trace_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_error_on_recompile_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_flat_name_to_original_fqn_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_fn_hasattr__name__1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_fn_hasattr__name__2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_fn_hasattr__name__3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_fold_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_frozen_dict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_frozenset_torch_func_contains_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_funcname_cache_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_function_annotation_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_generate_tensor_from_list_of_numpy_primitive_type_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_generate_trivial_abstract_impl_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_get_attr_function_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_get_cache_entry_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_get_custom_tensor_attribute_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_get_device_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_get_instruction_source_311_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_getattr_dict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_getset_descriptor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_grad_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_grad_non_none_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_grad_none_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_grad_state_mutated_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_graph_break_compilation_metrics_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_graph_break_compilation_metrics_on_failure_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_graph_break_correctly_when_passing_numpy_ndarray_to_torch_function_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_guard_failure_fn2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_guard_failure_fn_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_guard_failure_fn_shape_control_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_guard_failure_fn_tensor_iter_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_guard_function_builder_with_cse_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_guard_size_oblivious_backed_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_guard_size_oblivious_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_guard_sym_node_fstring_when_used_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_guards_cse_pass_multiple_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_guards_cse_pass_single_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_guards_strip_function_call_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_hasattr_nn_module_guard_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_hash_getitem_slice_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_id_guarded_module_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_id_guarded_object_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_id_of_nn_module_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_id_tensor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_if_cond_nn_mod1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_if_cond_nn_mod2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_if_cond_nn_mod3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_if_cond_user_defined_object2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_if_cond_user_defined_object3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_if_cond_user_defined_object_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inference_mode_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inline_closure_not_loaded_by_parent_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inline_dict_function_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inline_dict_function_passed_as_arg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inline_dict_mutation_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inline_func_jump_on_tensor_condition_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inline_list_mutation_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inline_local_dict_clear_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inline_module_attr_dict_clear_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inline_user_defined_dict_attr_clear_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inplace_desugaring_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inplace_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inplace_param_update_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inplace_view_on_graph_input_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_input_set_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inspect_signature_bind_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_inspect_signature_bind_non_user_function_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_int_int_comparisons_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_int_list_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_int_neg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_int_shape_binops_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_int_shape_comparisons_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_int_shape_inplace_binops_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_intermediary_tensor_grad_access_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_interpolate_propagate_real_tensors_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_invalid_args_builtin_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_is_compiling_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_is_floating_point2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_is_floating_point_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_is_tensor2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_is_tensor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_is_tensor_like2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_is_tensor_like_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_item_changes_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_item_changes_new_shape_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_item_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_iter_set_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_iter_type_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_itertools_accumulate_symint_default_sum_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_itertools_accumulate_tensors_builtins_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_itertools_accumulate_tensors_default_sum_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_itertools_accumulate_tensors_kwargs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_itertools_accumulate_tensors_user_defined_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_itertools_groupby_pure_python_default_identify_func_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_itertools_groupby_pure_python_key_func_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_itertools_infinite_count_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_itertools_infinite_cycle_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_itertools_infinite_repeat_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_itertools_infinite_repeat_mutation_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_itertools_repeat_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_large_reduction_list_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_linear_module_free_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_list_append_return_none_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_list_hasattr1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_list_hasattr2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_list_iadd_side_effect_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_list_iadd_with_shape_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_list_iterator_contains_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_list_mul_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_list_slice_mul_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_listcomp_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_load_fast_and_clear_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_mandelbrot_numpy_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_map_side_effects_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_map_with_quantization_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_mark_dynamic_with_ranges_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_mark_static_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_matmul1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_min_max_over_iterable_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_module_complex_iter_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_module_deepcopy_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_module_dunder_dict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_module_not_callable_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_named_parameters_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_namedtuple1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_namedtuple2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_namedtuple3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nan_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nested_closure_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nested_closure_mutation_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nested_function_resuming_with_correct_globals_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nested_optimize_decorator_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nested_optimize_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nested_optimize_run_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nested_sequential_try_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nested_sequential_try_with_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nested_sequential_try_with_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nested_sequential_with_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nested_wraps_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_new_with_int_list_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nn_functional_reduction_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nn_module_getattr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nn_module_getattribute_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nn_sequential_invocation_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nn_sequential_invocation_reposition_indices_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_no_error_on_nested_fx_trace_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_no_guard_for_unused_sym_node_fstring_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_no_raise_guard_partial_constraint_across_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_no_raise_guard_partial_constraint_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_non_pt2_compliant_ops_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_nonzero_static_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_not_dynamic_scope_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numel_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_array_of_arrays_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_as_global_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_fallback_on_eager_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_force_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_gt_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_int_constant_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_iter_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_min_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_ndarray_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_ndarray_graph_break_with_multiple_outputs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_ndarray_works_with_builtin_function_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_no_raise_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_non_torch_dtype_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_random_config_to_numpy_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_readonly_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_recompilation_scalar_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_size_attr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_subdtype_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_take_along_axis_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_tolist_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_torch_operators_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_ufunc_out_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_ufunc_out_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_unique_f16_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_variable_isinstance_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_numpy_with_builtin_type_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_object_classmethod_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_object_setattr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_object_staticmethod_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_onnx_shape_as_tensor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_optimize_on_module_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_optree_graph_break_message_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_ordered_dict_alias_reconstruct_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_ordered_dict_move_to_end_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_os_environ_get_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_os_environ_set_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_out_variant_custom_op_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_out_variants_with_resizing_on_graph_inputs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_out_variants_with_resizing_on_graph_inputs_with_dynamic1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_out_variants_with_resizing_on_graph_inputs_with_dynamic_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_outside_linear_module_free_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_packaging_version_parse_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_pair_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_param_shape_binops_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_parameter_free_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_parsing_sdpa_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_patched_builtin_functions_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_pt2_compliant_ops_are_allowed_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_pt2_compliant_overload_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_pure_python_accumulate_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_py_guards_mark_dynamic_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_python_slice_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_raise_guard_full_constraint_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_raise_guard_indirect_full_constraint_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_raise_guard_partial_constraint_across_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_raise_guard_partial_constraint_no_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_raise_on_backend_error_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_raises_importerror1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_raises_importerror2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_raises_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_rand_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_range_input_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_range_with_shape_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_real_imag_tensor_attribute_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_recompile_message_on_parameter_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_recompile_on_global_state_change_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_reconstruct_set_across_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_recursive_inline_list_mutation_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_recursive_tensor_attribute_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_release_input_memory_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_release_module_memory_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_release_scope_memory_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_repeat_interleave_graphbreaks_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_repro_graph_breaks_in__get_item_by_idx_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_restore_graphstate_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_return_dict_with_graph_break_and_update_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_return_nested_function_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_runtime_assert_replacement_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_sample_input_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_scalar_tensor_is_equivalent_to_int_list_argument_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_scalar_tensor_is_equivalent_to_symint_argument_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_scalar_tensor_is_equivalent_to_symint_list_argument_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_sequential_module_free_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_set_aliasing_recompiles_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_set_custom_tensor_attribute_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_setattr_mutation1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_setattr_mutation2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_setattr_mutation3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_and_tuple_equality_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_env_equal_constructor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_env_equal_create_symbolic_sizes_strides_storage_offset_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_env_equal_empty_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_env_equal_evaluate_expr_divisible_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_env_equal_evaluate_expr_refinement_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_env_equal_evaluate_expr_replacement_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_env_equal_runtime_assert_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_env_equal_unbacked_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_env_no_recording_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_env_recorded_function_fallback_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_int_comparisons_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_int_inplace_binops_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_type_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_shape_unpack_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_side_effects_codegen_update_mutated_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_simple_set_usage_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_size_dim_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_size_input_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_slice_input_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_source_non_input_grad_access_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_storage_return_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_str_format_assert1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_str_format_assert2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_str_format_return1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_str_format_return2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_stride_dim_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_super_after_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_super_calling_with_metaclass_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_sym_constrain_range_on_replaced_unbacked_symbol_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_symint_as_device_kwarg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_symint_as_device_kwarg_multi_gpu_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_symint_as_device_kwarg_non_strict_export_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_sys_modules_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tagging_tensors_mix_used_unused_structure_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tagging_tensors_simple_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_build_list_unpack_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_ctor_list_of_tensor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_data_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_dict1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_dict2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_dict3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_dot_grad_no_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_hasattr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_interacts_with_numpy_ndarray_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_is_contiguous_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_item_capture_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_item_no_capture_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_iter_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_layout_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tensor_types_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tolist_0d_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tolist_1d_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tolist_float_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tolist_kd_dynamic_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tolist_kd_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tolist_scalar_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_top_package_import_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_check_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_check_is_size_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_check_symbolic_shape_rel_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_compile_ctx_on_forward_and_training_step_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_cuda_is_available_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_cudnn_is_acceptable_bad_inputs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_cudnn_is_acceptable_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_device_python_type_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_distributions_lazy_property_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_dtype_python_type_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_dynamo_codegen_pow_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_generator_set_state_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_guards_stack_frame_register_inlining_deep_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_guards_stack_frame_register_inlining_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_nn_parameter_isinstance_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_objects_as_keys_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_package_working_with_trace_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_seed_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_size_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_size_numel_dynamic_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_size_numel_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_torch_variable_hasattr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_trace_ndarray_frame_2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_trace_ndarray_frame_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tracing_nested_py_tree_dicts_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tracing_nested_py_tree_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tracing_nested_py_tree_mixed_all_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tracing_nested_py_tree_tuples_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tracing_py_tree_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tracing_py_tree_tensor_subclass_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tracing_tree_map_only_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tuple_from_tuple_iter_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tuple_hasattr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tuple_iadd_with_shape_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tuple_mul_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_tuple_mul_with_shape_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_type_copy_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_typing_dict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_typing_typevar_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_typing_union_and_optional_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_typing_variable_isinstance_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_unbacked_auto_functionalize_op_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_unbacked_symint_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_unhandled_exception_in_dynamo2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_unhandled_exception_in_dynamo_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_unpack4_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_unpack5_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_unpack_tensor_shape_mismatch_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_update_locals_and_stack_uses_shared_cache_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_user_defined_binop_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_user_defined_class_name_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_user_defined_class_python_type_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_user_defined_iter_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_user_defined_setattr1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_user_defined_setattr2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_user_function_variable_supports_enum_argument_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_user_function_variable_supports_function_argument_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_user_function_variable_supports_type_abcmeta_argument_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_user_getattr1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_user_getattr2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_user_getattribute_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_user_property_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_usr_cls_classmethod_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_usr_cls_staticmethod_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_validate_outputs_unbacked_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_variable_access_in_exception_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_variable_tracker_recursively_contains_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_version_ci_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_with_builtin_type_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_write_to_closures_in_inlining_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_yield_from_in_a_loop_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_yield_from_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_yield_from_user_stop_iteration_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_yield_gen_and_from_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesMiscTests::test_yield_send_to_subgenerator_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_T_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_add__inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_add_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_addcdiv__inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_addcdiv_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_build_list_unpack_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_call_dict1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_call_dict2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_call_dict3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_call_dict4_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_call_dict5_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_callable_builtin_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_callable_class_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_callable_lambda_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_callable_list_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_callable_torch_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_chunks1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_class_dict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_cls_eq_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_cls_hasattr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_cls_is_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_compare_constant_and_tensor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_complex_closure_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_const_tuple_add1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_const_tuple_add2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_constant1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_constant2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_constant3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_constant4_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_context_wrapping_nested_functions_no_closure_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_cublas_allow_tf32_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_custom_dict_kwargs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_default_dict_closure_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_default_dict_constr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_default_dict_dict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_default_dict_lambda_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_default_dict_list_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_default_dict_set_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_default_dict_tuple_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_del_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_deque_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_device_constant_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_device_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_dict_copy_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_dict_fromkeys_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_dict_id_guard_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_dict_keys_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_dict_kwargs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_dict_mutable_map_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_dict_ops_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_dict_param_keys_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_dict_sorted_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_dict_tuple_lazy_guard_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_dict_update_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_dict_values_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_distributed_is_available_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_distributed_is_initialized_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_dtype_compare_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_dtype_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_elipsis_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_finfo_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_flat_param_same_storage_size_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_float_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_fn_with_self_set_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_fstrings1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_fstrings2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_fstrings3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_fstrings4_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_fstrings5_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_fstrings6_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_funcdef_closure_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_functools_partial_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_get_autocast_gpu_dtype_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_get_calculate_correct_fan_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_get_default_dtype_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_get_device_properties_tensor_device_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_get_privateuse1_name_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_globalfn_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_globalmodule_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_globalvar_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_import1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_in_not_in_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_index_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_indexed_range_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_indirect1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_indirect2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_indirect3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_inline_jit__unwrap_optional_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_inline_jit_annotations_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_inline_lru_cache_fn_with_default_args_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_inline_script_if_tracing_fn_with_default_args_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_inline_softmax_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_inline_with_default_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_inner_function_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_is_any_autocast_enabled_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_is_complex_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_is_contiguous_frame_counts_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_is_contiguous_memory_format_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_is_floating_point_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_is_fx_tracing_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_is_in_onnx_export_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_is_integer_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_is_not_null_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_is_quantized_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_is_sparse_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_islice_chain_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_itertools_chain_from_iterable_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_itertools_chain_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_itertools_combinations_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_itertools_product_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_jit_annotate_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_len_constant_dict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_len_constant_list_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_len_constant_misc_iterables_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_len_tensor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_list_add_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_list_add_then_mutate_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_list_clear_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_list_compare_polyfill_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_list_convert_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_list_expand_lhs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_list_index_with_constant_tensor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_list_reversed_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_list_slice_assignment_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_list_sorted1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_list_sorted2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_list_truth_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_listarg1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_listarg2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_listarg3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_listarg4_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_listarg5_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_load_global_bool_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_mT_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_manual_seed_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_map_sum_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_math_radians_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_mean_sum_np_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_methodcall1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_methodcall2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_methodcall3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_min_max_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_module_constant_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_namedtuple_defaults_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_namedtuple_hasattr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_namedtuple_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_namedtuple_user_methods_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_ndarray_builtin_functions_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_ndarray_method_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_ndarray_methods_returning_scalar_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_ndarray_reshape_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_ndarray_transpose_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_ndim_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_no_recompile_inner_function_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_no_recompile_inner_lambda_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_non_inlined_closure_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_not_list_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_np_constant_collections_as_input_int_or_float_float_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_np_constant_collections_as_input_int_or_float_int_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_np_constant_collections_guards_float_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_np_constant_collections_guards_int_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_np_finfo_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_np_iinfo_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_number_method_method_as_integer_ratio_num_type0_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_number_method_method_as_integer_ratio_num_type3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_number_method_method_bit_length_num_type1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_number_method_method_conjugate_num_type2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_number_method_method_conjugate_num_type4_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_number_method_method_hex_num_type5_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_number_method_method_is_integer_num_type6_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_numpy_attributes_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_numpy_dtype_argument_to_function_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_numpy_dtype_call_in_function_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_numpy_fft_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_numpy_linalg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_numpy_meshgrid_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_numpy_random_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_numpy_size_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_obj_eq_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_obj_is_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_ordered_dict_kwargs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partial_across_graph_break_uninvoked_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_as_input_UDF_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_as_input_partials_lambda_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_as_input_partials_mod_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_graph_break_reconstruct_args_and_kwargs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_graph_break_reconstruct_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_graph_break_reconstruct_mix_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_graph_break_reconstruct_mix_no_source_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___annotations___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___builtins___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___call___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___class___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___closure___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___code___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___defaults___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___delattr___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___dict___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___dir___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___doc___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___eq___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___format___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___ge___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___get___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___getattribute___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___globals___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___gt___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___hash___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___init___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___init_subclass___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___kwdefaults___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___le___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___lt___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___module___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___name___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___ne___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___new___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___qualname___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___reduce___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___reduce_ex___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___repr___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___setattr___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___sizeof___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___str___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr___subclasshook___inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr_args_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr_func_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_attr_keywords_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_hasattr_set_attr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_lambda_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_recompilation_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_torch_op_arg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_torch_op_kwarg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_udf_arg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_udf_kwarg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_udf_kwarg_method_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_partials_udf_kwarg_module_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_pop_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_pos_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_pow_int_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_promote_types_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_rand_inlined_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_rand_tensor_partial_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_range1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_range2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_range_length_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_range_with_index_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_range_with_slice_index_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_reduce_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_reduce_with_initial_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_reduce_with_none_initial_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_reduce_with_single_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_reduce_with_single_with_initial_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_return_dict2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_return_dict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_return_multiple_numpy_ndarray_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_return_numpy_ndarray_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_return_tuple1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_return_tuple2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_set_contains_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_set_difference_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_set_intersection_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_set_isdisjoint_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_set_keys_view_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_set_union_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_set_update_bytecode_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_set_update_list_with_duplicated_items_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_shape1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_shape2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_slice1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_slice2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_slice3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_slice4_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_slice5_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_slice6_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_sliced_range_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_startswith_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_sum_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_sum_shortcut_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_sum_shortcut_with_start_arg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_sum_shortcut_with_start_kwarg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_sum_with_start_arg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_sum_with_start_kwarg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_symbool_to_int_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tensor_element_size_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tensor_is_complex_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tensor_len_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tensor_new_with_shape_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tensor_new_with_size_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tensor_size_indexed_by_symint_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tensor_type2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tensor_type3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tensor_type4_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tensor_type5_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tensor_type_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_to_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_torch_distributions_functions_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_torch_from_numpy_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_torch_size_hasattr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_transpose_for_scores_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_truth_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tuple1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tuple2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tuple_contains_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tuple_iadd_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_tuple_sorted_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_unary_fold_op_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_unary_fold_op_seq_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_unpack1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_unpack2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_unpack3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_unpack_ex1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_unpack_ex2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_unpack_ex3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_unpack_mutable_map_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_viamethod_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_viatorch_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFunctionTests::test_zip_longest_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_access_by_keys_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_basicmodule1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_basicmodule2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_call_fn_with_non_const_inputs_safe_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_cfgmod_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_children_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_constloop_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_conv_call_forward_directly_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_conv_call_super_forward_directly_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_conv_transpose_call_forward_directly_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_conv_transpose_call_super_forward_directly_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_densenet_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_enumvalues_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_fnmember_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_fnmembercmp1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_fnmembercmp2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_forward_directly_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_generation_tag_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_hasattr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_intarg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_iseval1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_iseval2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_isnonelayer_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_istraining1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_istraining2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_layerlist_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_lazy_module1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_lazy_module2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_lazy_module3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_lazy_module4_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_lazy_module5_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_lazy_module6_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_lazy_module7_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_lazy_module_kwargs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_lazy_module_no_cls_to_become_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_module_attribute_precedence_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_module_call_module_with_static_forward_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_module_class_method_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_module_comparison_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_module_forward_has_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_module_guard_name_is_valid_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_module_name_string_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_module_property_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_module_static_method_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_moduledict_custom_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_moduledict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_modulelist_custom_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_modulelist_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_modulelist_nested_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_modulemethod1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_modulemethod2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_named_children_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_nn_moduledict_contains_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_parameterdict_custom_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_parameterdict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_parameters1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_parameters2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_parameters3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_parameters4_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_parameters5_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_self_mutating1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_seq_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_sequential_with_duplicated_module2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_sequential_with_duplicated_module_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_simple_torch_function_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_stringmember_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_submodules1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_submodules2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_super1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_super2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_super_class_method_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_tensorlist_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_torch_function_with_closure_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_unsupportedmethod_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_unsupportedmodule_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesNNModuleTests::test_viamodulecall_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_access_module_attr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_capture_constants_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_capture_global_num_adds_guard_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_capture_global_num_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_capture_input_num_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_capture_numpy_number_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_capture_tracked_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_capture_tracked_nested_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_capture_untracked_global_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_capture_untracked_global_nested_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_capture_untracked_nonlocal_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_capture_value_created_in_subgraph_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_cond_branches_no_arguments_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_cond_branches_no_arguments_no_closure_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_cond_free_variable_in_both_branches_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_cond_graph_break_in_one_branch_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_cond_pytree_operands_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_cond_pytree_operands_with_non_tensor_leaves_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_cond_side_effect_in_one_branches_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_cond_source_fn_stack_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_cond_subgraph_name_is_valid_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_cond_with_constant_pred_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_enum_arg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_error_message_sane_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_fallback_on_graph_break_complicated_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_fallback_on_graph_break_simple_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_fallback_on_python_primitives_output_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_flat_list_output_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_fn_with_kwargs_in_torch_ops_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_freevars_as_inputs_to_wrap_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_grad_source_fn_stack_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_hooks_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_inlined_functions_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_internal_nonlocal_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_lift_tensor_constant_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_make_closure_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_map_example_value_metadata_consistent_with_eager_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_map_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_map_kwargs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_map_lowers_to_graph_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_map_multi_return_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_map_pytree_return_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_map_side_effect_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_map_source_fn_stack_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_map_subgraph_name_is_valid_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_map_symint_input_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_modules_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_nested_tuple_output_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_nested_wrap_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_no_freevars_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_output_with_dict_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_register_mode_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_register_subclass_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_return_captured_var_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_return_captured_var_used_multiple_times_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_return_captured_vars_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_same_freevar_twice_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_del_existing_attr_global_module_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_del_existing_attr_global_obj_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_del_existing_attr_nonlocal_module_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_del_existing_attr_nonlocal_obj_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_in_body_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_local_list_append_no_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_mutate_global_list_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_mutate_global_num_builtin_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_mutate_global_num_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_mutate_global_tensor_builtin_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_mutate_global_tensor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_mutate_nonlocal_num_builtin_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_mutate_nonlocal_num_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_mutate_nonlocal_tensor_builtin_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_mutate_nonlocal_tensor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_nested_nonlocal_list_append_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_nonlocal_list_append_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_set_existing_attr_global_module_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_set_existing_attr_global_obj_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_set_existing_attr_nonlocal_module_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_set_existing_attr_nonlocal_obj_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_set_new_attr_global_module_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_set_new_attr_global_obj_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_set_new_attr_nonlocal_module_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_side_effect_set_new_attr_nonlocal_obj_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_symint_input_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_vmap_multiply_scalar_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_vmap_source_fn_stack_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_all_kwarg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_allow_local_assign_in_body_fn_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_kwarg_default_else_branch_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_kwarg_default_if_branch_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_kwarg_default_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_kwarg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_kwarg_int_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_kwarg_only_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_kwarg_recompile_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_pytree_args_nested_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_pytree_args_not_const_symint_tensor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_pytree_args_with_symint_constant_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_pytree_kwargs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_source_fn_stack_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesHigherOrderOpTests::test_wrap_subgraph_name_is_valid_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_functional_call_disable_capture_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_functional_call_disable_inline_nn_module_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_functional_call_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_functional_call_sequential_params_and_buffers_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_capture_tensor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_closure_scalar_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_disable_capture_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_fn_with_kwargs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_freevar_python_scalar_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_freevar_tensor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_has_aux_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_non_tensor_input_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_over_grad_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_pytree_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_recompile_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_two_tensor_all_grad_has_aux_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_two_tensor_has_aux_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_with_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_grad_with_side_effect_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_hessian_argnums_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_hessian_disable_capture_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_hessian_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jacfwd_disable_capture_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jacfwd_has_aux_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jacfwd_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jacfwd_randomness_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jacfwd_two_tensors_argnums_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jacrev_disable_capture_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jacrev_has_aux_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jacrev_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jacrev_two_tensors_argnums_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jvp_disable_capture_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jvp_freevar_python_scalar_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jvp_freevar_tensor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jvp_has_aux_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jvp_jvp_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jvp_simple_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jvp_two_tensors_disable_enable_disable_grad_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jvp_two_tensors_disable_grad_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_jvp_two_tensors_has_aux_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_linearize_disable_capture_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_linearize_jvp_fn_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vjp_disable_capture_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vjp_has_aux_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vjp_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vjp_multiple_outputs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vjp_multiple_outputs_python_struct_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_disable_capture_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_free_const_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_free_tensor_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_get_wrapped_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_kwargs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_multiple_invocation_in_dims_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_multiple_invocation_out_dims_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_multiple_outputs_diff_dims_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_multiple_outputs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_multiple_outputs_out_dims_tuple_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_new_tensor_implicit_via_op_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_new_tensor_in_body_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_new_tensor_unused_in_body_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_over_vmap_captured_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_over_vmap_two_inputs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_previous_illegal_op_no_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_pytree_inputs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_recompile_different_config_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_recompile_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_recompile_same_config_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_recompile_with_randomness_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_side_effects_append_input_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_side_effects_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_two_inputs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_two_inputs_tuple_in_dims_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_with_conditional_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_with_graph_break_2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_with_graph_break_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesFuncTorchHigherOrderOpTests::test_vmap_with_graph_break_lambda_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_LSTM_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_alias_inputs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_aot_autograd_expand_mutation_backwards_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_aot_autograd_expand_mutation_error_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_aot_autograd_expand_mutation_functionalizes_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_aot_autograd_raises_invalid_leaf_set_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_aot_export_joint_simple_repro_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_aot_grad_mode_mutation_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_aot_sequence_nr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_arg_dupe_via_dynamo_recompiles_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_arg_dupe_via_dynamo_recompiles_many_args_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_arg_dupe_via_dynamo_recompiles_many_args_param_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_arg_dupe_via_dynamo_recompiles_many_args_param_non_tensor_arg_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_arg_dupe_via_dynamo_recompiles_many_args_param_non_tensor_arg_list_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_arg_dupe_via_dynamo_recompiles_many_with_global_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_call_fn_with_non_const_inputs_aot_safe_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_call_fn_with_non_const_inputs_aot_unsafe_control_flow_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_call_fn_with_non_const_inputs_aot_unsafe_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_data_ptr_access_copy_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_data_ptr_access_fails_in_backward_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_data_ptr_access_fails_in_forward_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_donated_buffer1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_donated_buffer2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_donated_buffer3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_donated_buffer4_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_donated_buffer5_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_donated_buffer_with_retain_or_create_graph1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_donated_buffer_with_retain_or_create_graph2_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_donated_buffer_with_retain_or_create_graph3_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_donated_buffer_with_retain_or_create_graph4_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_double_backward_errors_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_eager_sequence_nr_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_grad_inputs_alias_inputs_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_multiple_aot_autograd_calls_dupe_args_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_mutation1_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_mutation_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_negative_testing_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_negative_testing_mutation_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_nn_parameter_construction_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_requires_grad_fake_via_dynamo_recompiles_inline_inbuilt_nn_modules, test/dynamo/test_inline_inbuilt_nn_modules.py::InlineInbuiltNNModulesAotAutogradFallbackTests::test_split_with_sizes_aot_autograd_cleans_up_traceback_meta_inline_inbuilt_nn_modules 2024-08-06T21:54:20.7050953Z 2024-08-06T21:54:24.4392082Z Running inductor/test_inductor_utils 1/1 ... [2024-08-06 21:54:24.438720] 2024-08-06T21:54:24.4396614Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_inductor_utils.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:54:24.439240] 2024-08-06T21:54:28.9626316Z 2024-08-06T21:54:28.9627475Z inductor/test_inductor_utils 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_inductor_utils_1.1_629e1c9620efd0bf_.log 2024-08-06T21:54:28.9629140Z Running 2 items in this shard: test/inductor/test_inductor_utils.py::TestBench::test_do_bench, test/inductor/test_inductor_utils.py::TestBench::test_do_bench_using_profiling 2024-08-06T21:54:28.9630002Z 2024-08-06T21:54:32.6090305Z Running torch_np/test_unary_ufuncs 1/1 ... [2024-08-06 21:54:32.608456] 2024-08-06T21:54:32.6095196Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'torch_np/test_unary_ufuncs.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:54:32.608958] 2024-08-06T21:54:37.0321049Z 2024-08-06T21:54:37.0322145Z torch_np/test_unary_ufuncs 1/1 was successful, full logs can be found in artifacts with path test/test-reports/torch_np.test_unary_ufuncs_1.1_fd5722761f1a66f9_.log 2024-08-06T21:54:37.0335963Z Running 42 items in this shard: test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_absolute, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_arccos, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_arccosh, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_arcsin, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_arcsinh, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_arctan, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_arctanh, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_cbrt, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_ceil, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_conjugate, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_cos, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_cosh, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_deg2rad, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_degrees, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_exp, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_exp2, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_expm1, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_fabs, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_floor, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_isfinite, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_isinf, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_isnan, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_log, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_log10, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_log1p, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_log2, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_logical_not, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_negative, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_positive, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_rad2deg, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_radians, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_reciprocal, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_rint, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_sign, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_signbit, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_sin, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_sinh, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_sqrt, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_square, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_tan, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_tanh, test/torch_np/test_unary_ufuncs.py::TestUnaryUfuncs::test_trunc 2024-08-06T21:54:37.0348859Z 2024-08-06T21:54:40.7334762Z Running inductor/test_extension_backend 1/1 ... [2024-08-06 21:54:40.733005] 2024-08-06T21:54:40.7338458Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_extension_backend.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:54:40.733462] 2024-08-06T21:55:12.7137221Z 2024-08-06T21:55:12.7138393Z inductor/test_extension_backend 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_extension_backend_1.1_6f606e5b2191a99d_.log 2024-08-06T21:55:12.7139851Z Running 1 items in this shard: test/inductor/test_extension_backend.py::ExtensionBackendTests::test_open_device_registration 2024-08-06T21:55:12.7140525Z 2024-08-06T21:55:16.5453666Z Running export/test_export_nonstrict 1/1 ... [2024-08-06 21:55:16.544841] 2024-08-06T21:55:16.5457175Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'export/test_export_nonstrict.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:55:16.545328] 2024-08-06T21:55:57.2849161Z 2024-08-06T21:55:57.2851224Z export/test_export_nonstrict 1/1 was successful, full logs can be found in artifacts with path test/test-reports/export.test_export_nonstrict_1.1_e932317b1a2734d9_.log 2024-08-06T21:55:57.2953084Z Running 179 items in this shard: test/export/test_export_nonstrict.py::NonStrictExportTestDynamismExpression::test_export_assume_static_by_default_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestDynamismExpression::test_export_constraints_error_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestDynamismExpression::test_export_constraints_error_not_in_range_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestDynamismExpression::test_export_inline_constraints_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestDynamismExpression::test_export_slice_maxsize_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test__scaled_dot_product_flash_attention_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_allow_explicit_guards_as_runtime_asserts_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_args_type_checked_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_aten_lift_fresh_copy_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_automatic_constrain_size_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_basic_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_basic_non_strict_fake_tensor_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_basic_non_strict_real_tensor_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_buffer_util_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_check_is_size_error_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_check_specialized_int_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_checks_to_constrain_range_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_colon_parameter_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_compiling_state_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_cond_buffers_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_cond_with_module_stack_export_with_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_cond_with_module_stack_export_with_unflatten_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_constant_aliasing_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_constant_input_naming_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_constant_output_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_constrain_decomp_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_constrain_size_in_eager_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_constrain_size_with_constrain_value_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_constrain_size_with_various_cases_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_conv_dynamic_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_crop_like_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_cse_for_symint_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_custom_op_auto_functionalize_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_custom_op_auto_functionalize_pre_dispatch_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_custom_op_auto_warn_pre_dispatch_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_decomp_batch_norm_functional_predispatch_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_derived_dim_1_2_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_derived_dim_basic_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_derived_dim_integer_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_derived_dim_nested_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_derived_dim_out_of_order_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_derived_dim_out_of_order_repeat_derived_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_derived_dim_out_of_order_simplified_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_derived_dim_out_of_order_simplified_repeat_non_derived_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_derived_dim_repeat_derived_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_device_to_dynamic_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_device_to_mutation_float_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_device_to_mutation_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_device_to_static_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_dim_1_2_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_disable_forced_specializations_errors_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_disable_forced_specializations_ok_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_dynamic_shapes_builder_basic_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_dynamic_shapes_builder_kwargs_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_dynamic_shapes_builder_pytree_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_dynamic_shapes_spec_with_pytree_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_error_does_not_reference_eager_fallback_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_api_with_dynamic_shapes_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_as_backend_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_cond_preserve_torch_fn_for_subgraphs_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_cond_symbool_pred_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_decomps_dynamic_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_decomps_simple_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_dynamo_config_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_for_training_run_decomp_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_for_training_with_container_type_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_for_training_with_dynamic_shapes_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_for_training_with_mutation_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_func_with_default_kwargs_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_func_with_keyword_only_args_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_func_with_kwargs_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_func_with_pytree_kwargs_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_func_with_var_keyword_args_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_func_with_var_keyword_pytree_args_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_func_with_var_postional_args_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_graph_with_no_inputs_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_input_mutation_bug_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_input_mutation_dynamic_shape_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_input_mutation_static_shape_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_mod_constraints_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_predispatch_custom_ops_warnings_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_preserve_linear_at_aot_level_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_then_compile_tensor_ctor_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_with_fake_tensor_inputs_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_with_fake_tensor_inputs_on_cuda_devices_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_with_inline_constraints_complex_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_with_inline_constraints_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_export_with_wrong_inputs_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_external_call_non_strict_real_tensor_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_fake_inputs_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_fake_weights_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_float_conversion_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_fqn_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_if_functional_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_intermediate_shape_comp_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_issue_113041_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_keep_composite_ops_invalid_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_keep_composite_ops_linear_convd_for_training_ir_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_keep_composite_ops_linear_convd_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_layer_sharing_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_lazy_module_kwargs_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_lifted_constants_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_linear_conv_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_map_buffers_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_map_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_mixed_input_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_module_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_module_with_dict_container_inp_out_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_multiple_definitions_same_name_dim_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_nested_dynamic_shapes_spec_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_nested_module_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_nested_module_with_constant_buffer_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_nested_module_with_init_buffer_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_nested_module_with_parameter_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_nn_module_stack_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_nn_module_stack_shared_submodule_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_no_tensor_computation_2_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_no_tensor_computation_3_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_no_tensor_computation_4_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_no_tensor_computation_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_non_arg_name_dynamic_shapes_api_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_non_arg_name_dynamic_shapes_api_with_container_type_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_non_arg_name_dynamic_shapes_api_with_kwarg_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_non_persistent_buffer_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_non_strict_dynamic_shapes_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_non_strict_dynamic_shapes_suggested_fixes_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_nonstrict_retrace_preserves_metadata_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_nonzero_2_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_not_correct_dim_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_pad_sequence_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_param_util_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_placeholder_naming_collisions_hoo_subgraphs_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_placeholder_naming_collisions_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_predispatch_cond_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_predispatch_grad_wrappers_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_preserve_requires_grad_placeholders_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_preserve_shape_dynamism_for_unused_inputs_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_pytree_register_data_class_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_pytree_register_nested_data_class_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_raise_user_error_when_guard_on_data_dependent_operation_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_redundant_asserts_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_refine_dynamic_shapes_from_suggested_fixes_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_reshape_view_helper_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_retracable_ep_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_retrace_graph_level_meta_preservation_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_retrace_pre_autograd_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_run_decomposition_supports_user_input_mutation_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_runtime_assert_for_prim_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_runtime_assert_for_prm_str_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_runtime_assert_with_size_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_set_grad_empty_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_setgrad_lifted_tensor_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_simple_export_for_training_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_slice_with_floordiv_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_solver_unsupported_sympy_function_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_specialize_derived_dim_roots_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_split_const_gm_with_lifted_constants_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_stack_trace_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_state_primitives_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_state_tensors_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_static_dim_constraints_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_suggested_fixes_for_data_dependent_errors_basic_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_suggested_fixes_for_data_dependent_errors_puzzlers_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_suggested_fixes_new_roots_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_sym_sqrt_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_symint_tensor_return_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_tensor_attribute_zero_args_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_to_module_with_mutated_buffer_multiple_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_to_module_with_mutated_buffer_multiple_update_sub_later_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_to_module_with_mutated_buffer_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_torch_check_eq_commutativity_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_torch_fn_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_trace_under_fake_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_train_eval_on_exported_preautograd_module_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_unbacked_deferred_runtime_retrace_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_unbacked_slice_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_unflatten_asserts_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_unused_aliases_non_strict, test/export/test_export_nonstrict.py::NonStrictExportTestExport::test_user_input_and_buffer_mutation_non_strict 2024-08-06T21:55:57.3048180Z 2024-08-06T21:56:01.0917752Z Running inductor/test_triton_extension_backend 1/1 ... [2024-08-06 21:56:01.091288] 2024-08-06T21:56:01.0921257Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_triton_extension_backend.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:56:01.091751] 2024-08-06T21:56:11.3780317Z 2024-08-06T21:56:11.3782332Z inductor/test_triton_extension_backend 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_triton_extension_backend_1.1_846fcb8ddfbb5824_.log 2024-08-06T21:56:11.3785368Z Running 1 items in this shard: test/inductor/test_triton_extension_backend.py::TritonExtensionBackendTests::test_open_device_registration 2024-08-06T21:56:11.3786817Z 2024-08-06T21:56:15.1389218Z Running inductor/test_triton_kernels 1/1 ... [2024-08-06 21:56:15.138372] 2024-08-06T21:56:15.1392600Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_triton_kernels.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:56:15.138840] 2024-08-06T21:56:19.8119929Z 2024-08-06T21:56:19.8124099Z inductor/test_triton_kernels 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_triton_kernels_1.1_2aed89d432dfb97b_.log 2024-08-06T21:56:19.8235680Z Running 189 items in this shard: test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_False_backend_aot_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_False_backend_aot_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_False_backend_aot_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_False_backend_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_False_backend_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_False_backend_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_False_backend_inductor_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_False_backend_inductor_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_False_backend_inductor_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_True_backend_aot_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_True_backend_aot_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_True_backend_aot_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_True_backend_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_True_backend_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_True_backend_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_True_backend_inductor_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_True_backend_inductor_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_False_dynamic_True_backend_inductor_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_False_backend_aot_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_False_backend_aot_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_False_backend_aot_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_False_backend_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_False_backend_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_False_backend_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_False_backend_inductor_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_False_backend_inductor_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_False_backend_inductor_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_True_backend_aot_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_True_backend_aot_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_True_backend_aot_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_True_backend_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_True_backend_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_True_backend_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_True_backend_inductor_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_True_backend_inductor_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_2d_autotune_grad_True_dynamic_True_backend_inductor_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_False_backend_aot_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_False_backend_aot_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_False_backend_aot_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_False_backend_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_False_backend_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_False_backend_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_False_backend_inductor_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_False_backend_inductor_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_False_backend_inductor_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_True_backend_aot_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_True_backend_aot_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_True_backend_aot_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_True_backend_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_True_backend_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_True_backend_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_True_backend_inductor_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_True_backend_inductor_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_False_dynamic_True_backend_inductor_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_False_backend_aot_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_False_backend_aot_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_False_backend_aot_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_False_backend_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_False_backend_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_False_backend_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_False_backend_inductor_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_False_backend_inductor_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_False_backend_inductor_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_True_backend_aot_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_True_backend_aot_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_True_backend_aot_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_True_backend_eager_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_True_backend_eager_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_True_backend_eager_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_True_backend_inductor_grid_type_1, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_True_backend_inductor_grid_type_2, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_grad_True_dynamic_True_backend_inductor_grid_type_3, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_with_unsupported_args_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_with_unsupported_args_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_autotune_with_unsupported_args_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_caching, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_caching_duplicate, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_constants, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_dependancies, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_different_shapes_size_16_dynamic_False, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_different_shapes_size_16_dynamic_True, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_different_shapes_size_4_dynamic_False, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_different_shapes_size_4_dynamic_True, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_equal_to_1_arg_dynamic_False, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_equal_to_1_arg_dynamic_True, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_equal_to_1_float_arg_dynamic_False, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_equal_to_1_float_arg_dynamic_True, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_fallback, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_functionalize, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_higher_order_func, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_inner_triton_function_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_inner_triton_function_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_inner_triton_function_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_inputs_buffer_reuse, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_matmul_tracking, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_multi_kernel_grad_False, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_multi_kernel_grad_True, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_multiple_outputs_dynamic_False_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_multiple_outputs_dynamic_False_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_multiple_outputs_dynamic_False_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_multiple_outputs_dynamic_True_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_multiple_outputs_dynamic_True_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_multiple_outputs_dynamic_True_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_mutation_not_mark_dirty, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_mutation_type, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_native_grad_False_dynamic_False_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_native_grad_False_dynamic_False_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_native_grad_False_dynamic_False_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_native_grad_False_dynamic_True_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_native_grad_False_dynamic_True_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_native_grad_False_dynamic_True_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_native_grad_True_dynamic_False_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_native_grad_True_dynamic_False_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_native_grad_True_dynamic_False_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_native_grad_True_dynamic_True_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_native_grad_True_dynamic_True_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_native_grad_True_dynamic_True_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_no_clones_grad_False_dynamic_False, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_no_clones_grad_False_dynamic_True, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_no_clones_grad_True_dynamic_False, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_no_clones_grad_True_dynamic_True, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_num_ctas_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_num_ctas_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_num_ctas_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_out_of_order, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_reinplace_inplaceable_pass, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_reset_to_zero, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_slice_and_view_input, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_special_kwargs_with_autotune_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_special_kwargs_with_autotune_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_special_kwargs_with_autotune_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_special_kwargs_without_autotune_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_special_kwargs_without_autotune_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_special_kwargs_without_autotune_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_strided_input, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_strided_input_nonzero_offset, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_tracing_dynamic_False, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_tracing_dynamic_True, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_triton_dtype_dynamic_False_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_triton_dtype_dynamic_False_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_triton_dtype_dynamic_False_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_triton_dtype_dynamic_True_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_triton_dtype_dynamic_True_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_triton_dtype_dynamic_True_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_unbacked_shape_tensor_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_unbacked_shape_tensor_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_unbacked_shape_tensor_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_various_args, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_grad_option_grad_fn0_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_grad_option_grad_fn0_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_grad_option_grad_fn0_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_grad_option_grad_fn1_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_grad_option_grad_fn1_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_grad_option_grad_fn1_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_imported_symbol, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_imported_symbol_with_custom_name, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_kernel_param, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_views_dynamic_False_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_views_dynamic_False_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_views_dynamic_False_backend_inductor, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_views_dynamic_True_backend_aot_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_views_dynamic_True_backend_eager, test/inductor/test_triton_kernels.py::KernelTests::test_triton_kernel_with_views_dynamic_True_backend_inductor, test/inductor/test_triton_kernels.py::MutationTests::test_add_for_loop, test/inductor/test_triton_kernels.py::MutationTests::test_add_for_loop2, test/inductor/test_triton_kernels.py::MutationTests::test_add_nested_for_loop, test/inductor/test_triton_kernels.py::MutationTests::test_add_nested_for_loop_multi_return, test/inductor/test_triton_kernels.py::MutationTests::test_argmax, test/inductor/test_triton_kernels.py::MutationTests::test_cumsum, test/inductor/test_triton_kernels.py::MutationTests::test_fn_call_multi_return, test/inductor/test_triton_kernels.py::MutationTests::test_fn_call_one_return, test/inductor/test_triton_kernels.py::MutationTests::test_for_loop_arg, test/inductor/test_triton_kernels.py::MutationTests::test_for_loop_arg_2, test/inductor/test_triton_kernels.py::MutationTests::test_labels, test/inductor/test_triton_kernels.py::MutationTests::test_nested_cond_op_kernel, test/inductor/test_triton_kernels.py::MutationTests::test_out_of_order_kernel, test/inductor/test_triton_kernels.py::MutationTests::test_out_of_order_kernel_call, test/inductor/test_triton_kernels.py::MutationTests::test_reduce_sum, test/inductor/test_triton_kernels.py::MutationTests::test_triton_kernel_inference_mode, test/inductor/test_triton_kernels.py::MutationTests::test_while_loop, test/inductor/test_triton_kernels.py::CustomOpTests::test_add_kernel_autotuned_False_dynamic_False, test/inductor/test_triton_kernels.py::CustomOpTests::test_add_kernel_autotuned_False_dynamic_True, test/inductor/test_triton_kernels.py::CustomOpTests::test_add_kernel_autotuned_True_dynamic_False, test/inductor/test_triton_kernels.py::CustomOpTests::test_add_kernel_autotuned_True_dynamic_True, test/inductor/test_triton_kernels.py::CustomOpTests::test_capture_triton_meta, test/inductor/test_triton_kernels.py::CustomOpTests::test_capture_triton_special_kwargs_dynamic_False_autotune_False, test/inductor/test_triton_kernels.py::CustomOpTests::test_capture_triton_special_kwargs_dynamic_False_autotune_True, test/inductor/test_triton_kernels.py::CustomOpTests::test_capture_triton_special_kwargs_dynamic_True_autotune_False, test/inductor/test_triton_kernels.py::CustomOpTests::test_capture_triton_special_kwargs_dynamic_True_autotune_True 2024-08-06T21:56:19.8334104Z 2024-08-06T21:56:23.6073696Z Running dynamo/test_base_output 1/1 ... [2024-08-06 21:56:23.606899] 2024-08-06T21:56:23.6078052Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'dynamo/test_base_output.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:56:23.607397] 2024-08-06T21:56:27.4287903Z 2024-08-06T21:56:27.4289269Z dynamo/test_base_output 1/1 was successful, full logs can be found in artifacts with path test/test-reports/dynamo.test_base_output_1.1_d8394e509c5348f9_.log 2024-08-06T21:56:27.4292853Z Running 6 items in this shard: test/dynamo/test_base_output.py::TestBaseOutput::test_assign, test/dynamo/test_base_output.py::TestBaseOutput::test_create, test/dynamo/test_base_output.py::TestBaseOutput::test_getattr, test/dynamo/test_base_output.py::TestBaseOutput::test_getitem, test/dynamo/test_base_output.py::TestBaseOutput::test_index, test/dynamo/test_base_output.py::TestBaseOutput::test_tuple 2024-08-06T21:56:27.4294905Z 2024-08-06T21:56:31.0423418Z Running export/test_serdes 1/1 ... [2024-08-06 21:56:31.041889] 2024-08-06T21:56:31.0427318Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'export/test_serdes.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:56:31.042376] 2024-08-06T21:57:14.5375975Z 2024-08-06T21:57:14.5378189Z export/test_serdes 1/1 was successful, full logs can be found in artifacts with path test/test-reports/export.test_serdes_1.1_13f5b38669279a9b_.log 2024-08-06T21:57:14.5547616Z Running 179 items in this shard: test/export/test_serdes.py::SerDesExportTestDynamismExpression::test_export_assume_static_by_default_serdes, test/export/test_serdes.py::SerDesExportTestDynamismExpression::test_export_constraints_error_not_in_range_serdes, test/export/test_serdes.py::SerDesExportTestDynamismExpression::test_export_constraints_error_serdes, test/export/test_serdes.py::SerDesExportTestDynamismExpression::test_export_inline_constraints_serdes, test/export/test_serdes.py::SerDesExportTestDynamismExpression::test_export_slice_maxsize_serdes, test/export/test_serdes.py::SerDesExportTestExport::test__scaled_dot_product_flash_attention_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_allow_explicit_guards_as_runtime_asserts_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_args_type_checked_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_aten_lift_fresh_copy_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_automatic_constrain_size_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_basic_non_strict_fake_tensor_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_basic_non_strict_real_tensor_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_basic_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_buffer_util_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_check_is_size_error_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_check_specialized_int_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_checks_to_constrain_range_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_colon_parameter_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_compiling_state_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_cond_buffers_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_cond_with_module_stack_export_with_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_cond_with_module_stack_export_with_unflatten_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_constant_aliasing_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_constant_input_naming_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_constant_output_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_constrain_decomp_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_constrain_size_in_eager_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_constrain_size_with_constrain_value_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_constrain_size_with_various_cases_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_conv_dynamic_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_crop_like_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_cse_for_symint_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_custom_op_auto_functionalize_pre_dispatch_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_custom_op_auto_functionalize_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_custom_op_auto_warn_pre_dispatch_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_decomp_batch_norm_functional_predispatch_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_derived_dim_1_2_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_derived_dim_basic_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_derived_dim_integer_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_derived_dim_nested_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_derived_dim_out_of_order_repeat_derived_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_derived_dim_out_of_order_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_derived_dim_out_of_order_simplified_repeat_non_derived_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_derived_dim_out_of_order_simplified_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_derived_dim_repeat_derived_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_device_to_dynamic_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_device_to_mutation_float_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_device_to_mutation_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_device_to_static_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_dim_1_2_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_disable_forced_specializations_errors_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_disable_forced_specializations_ok_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_dynamic_shapes_builder_basic_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_dynamic_shapes_builder_kwargs_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_dynamic_shapes_builder_pytree_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_dynamic_shapes_spec_with_pytree_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_error_does_not_reference_eager_fallback_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_api_with_dynamic_shapes_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_as_backend_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_cond_preserve_torch_fn_for_subgraphs_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_cond_symbool_pred_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_decomps_dynamic_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_decomps_simple_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_dynamo_config_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_for_training_run_decomp_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_for_training_with_container_type_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_for_training_with_dynamic_shapes_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_for_training_with_mutation_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_func_with_default_kwargs_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_func_with_keyword_only_args_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_func_with_kwargs_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_func_with_pytree_kwargs_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_func_with_var_keyword_args_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_func_with_var_keyword_pytree_args_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_func_with_var_postional_args_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_graph_with_no_inputs_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_input_mutation_bug_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_input_mutation_dynamic_shape_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_input_mutation_static_shape_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_mod_constraints_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_predispatch_custom_ops_warnings_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_preserve_linear_at_aot_level_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_then_compile_tensor_ctor_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_with_fake_tensor_inputs_on_cuda_devices_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_with_fake_tensor_inputs_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_with_inline_constraints_complex_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_with_inline_constraints_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_export_with_wrong_inputs_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_external_call_non_strict_real_tensor_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_fake_inputs_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_fake_weights_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_float_conversion_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_fqn_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_if_functional_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_intermediate_shape_comp_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_issue_113041_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_keep_composite_ops_invalid_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_keep_composite_ops_linear_convd_for_training_ir_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_keep_composite_ops_linear_convd_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_layer_sharing_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_lazy_module_kwargs_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_lifted_constants_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_linear_conv_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_map_buffers_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_map_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_mixed_input_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_module_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_module_with_dict_container_inp_out_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_multiple_definitions_same_name_dim_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_nested_dynamic_shapes_spec_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_nested_module_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_nested_module_with_constant_buffer_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_nested_module_with_init_buffer_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_nested_module_with_parameter_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_nn_module_stack_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_nn_module_stack_shared_submodule_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_no_tensor_computation_2_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_no_tensor_computation_3_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_no_tensor_computation_4_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_no_tensor_computation_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_non_arg_name_dynamic_shapes_api_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_non_arg_name_dynamic_shapes_api_with_container_type_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_non_arg_name_dynamic_shapes_api_with_kwarg_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_non_persistent_buffer_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_non_strict_dynamic_shapes_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_non_strict_dynamic_shapes_suggested_fixes_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_nonstrict_retrace_preserves_metadata_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_nonzero_2_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_not_correct_dim_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_pad_sequence_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_param_util_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_placeholder_naming_collisions_hoo_subgraphs_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_placeholder_naming_collisions_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_predispatch_cond_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_predispatch_grad_wrappers_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_preserve_requires_grad_placeholders_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_preserve_shape_dynamism_for_unused_inputs_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_pytree_register_data_class_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_pytree_register_nested_data_class_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_raise_user_error_when_guard_on_data_dependent_operation_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_redundant_asserts_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_refine_dynamic_shapes_from_suggested_fixes_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_reshape_view_helper_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_retracable_ep_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_retrace_graph_level_meta_preservation_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_retrace_pre_autograd_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_run_decomposition_supports_user_input_mutation_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_runtime_assert_for_prim_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_runtime_assert_for_prm_str_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_runtime_assert_with_size_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_set_grad_empty_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_setgrad_lifted_tensor_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_simple_export_for_training_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_slice_with_floordiv_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_solver_unsupported_sympy_function_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_specialize_derived_dim_roots_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_split_const_gm_with_lifted_constants_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_stack_trace_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_state_primitives_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_state_tensors_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_static_dim_constraints_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_suggested_fixes_for_data_dependent_errors_basic_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_suggested_fixes_for_data_dependent_errors_puzzlers_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_suggested_fixes_new_roots_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_sym_sqrt_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_symint_tensor_return_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_tensor_attribute_zero_args_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_to_module_with_mutated_buffer_multiple_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_to_module_with_mutated_buffer_multiple_update_sub_later_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_to_module_with_mutated_buffer_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_torch_check_eq_commutativity_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_torch_fn_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_trace_under_fake_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_train_eval_on_exported_preautograd_module_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_unbacked_deferred_runtime_retrace_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_unbacked_slice_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_unflatten_asserts_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_unused_aliases_serdes, test/export/test_serdes.py::SerDesExportTestExport::test_user_input_and_buffer_mutation_serdes 2024-08-06T21:57:14.5714275Z 2024-08-06T21:57:18.2986893Z Running dynamo/test_interop 1/1 ... [2024-08-06 21:57:18.298133] 2024-08-06T21:57:18.2990029Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'dynamo/test_interop.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:57:18.298615] 2024-08-06T21:57:22.9224011Z 2024-08-06T21:57:22.9225719Z dynamo/test_interop 1/1 was successful, full logs can be found in artifacts with path test/test-reports/dynamo.test_interop_1.1_e7dd8e7c2ac12d8c_.log 2024-08-06T21:57:22.9227709Z Running 4 items in this shard: test/dynamo/test_interop.py::InteropTests::test_fx_fn, test/dynamo/test_interop.py::InteropTests::test_script_fn, test/dynamo/test_interop.py::InteropTests::test_trace_fn, test/dynamo/test_interop.py::InteropTests::test_vmap_in_graph 2024-08-06T21:57:22.9229023Z 2024-08-06T21:57:26.5878021Z Running inductor/test_compile_worker 1/1 ... [2024-08-06 21:57:26.587244] 2024-08-06T21:57:26.5881690Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_compile_worker.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:57:26.587754] 2024-08-06T21:57:43.5428123Z 2024-08-06T21:57:43.5429326Z inductor/test_compile_worker 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_compile_worker_1.1_20d1a87fa1171069_.log 2024-08-06T21:57:43.5431330Z Running 3 items in this shard: test/inductor/test_compile_worker.py::TestCompileWorker::test_basic_jobs, test/inductor/test_compile_worker.py::TestCompileWorker::test_crash, test/inductor/test_compile_worker.py::TestCompileWorker::test_exception 2024-08-06T21:57:43.5432613Z 2024-08-06T21:57:47.2917732Z Running dynamo/test_fx_passes_pre_grad 1/1 ... [2024-08-06 21:57:47.291263] 2024-08-06T21:57:47.2921124Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'dynamo/test_fx_passes_pre_grad.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:57:47.291726] 2024-08-06T21:57:53.3680546Z 2024-08-06T21:57:53.3682402Z dynamo/test_fx_passes_pre_grad 1/1 was successful, full logs can be found in artifacts with path test/test-reports/dynamo.test_fx_passes_pre_grad_1.1_7bc28a1b2f7f494f_.log 2024-08-06T21:57:53.3684338Z Running 1 items in this shard: test/dynamo/test_fx_passes_pre_grad.py::FxPassesPreGradTests::test_pass_execution_and_save 2024-08-06T21:57:53.3684984Z 2024-08-06T21:57:57.1255205Z Running inductor/test_flex_decoding 1/1 ... [2024-08-06 21:57:57.124894] 2024-08-06T21:57:57.1259136Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_flex_decoding.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:57:57.125403] 2024-08-06T21:58:01.6988585Z 2024-08-06T21:58:01.6990626Z inductor/test_flex_decoding 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_flex_decoding_1.1_cc23c085ddf11236_.log 2024-08-06T21:58:01.7168325Z Running 171 items in this shard: test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod0_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod0_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod0_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod0_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod1_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod1_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod1_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod1_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod2_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod2_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod2_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod2_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod3_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod3_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod3_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod3_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod4_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod4_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod4_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod4_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod5_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod5_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod5_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod5_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod6_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod6_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod6_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod6_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod7_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod7_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod7_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod7_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod8_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod8_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod8_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float16_score_mod8_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod0_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod0_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod0_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod0_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod1_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod1_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod1_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod1_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod2_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod2_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod2_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod2_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod3_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod3_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod3_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod3_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod4_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod4_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod4_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod4_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod5_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod5_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod5_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod5_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod6_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod6_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod6_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod6_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod7_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod7_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod7_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod7_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod8_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod8_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod8_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_builtin_score_mods_float32_score_mod8_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_bw_decoding_fails_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_captured_buffers_all_dims_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_captured_buffers_all_dims_float32, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_captured_buffers_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_captured_buffers_float32, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_captured_reduction_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_captured_scale_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_function_composition_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_function_composition_float32, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_load_from_bias_head_seq_batch_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_load_from_bias_seq_batch_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_load_from_bias_seq_only_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_logsumexp_correctness_float16_score_mod0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_logsumexp_correctness_float16_score_mod1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_logsumexp_correctness_float32_score_mod0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_logsumexp_correctness_float32_score_mod1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_logsumexp_is_not_fused, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_logsumexp_only_return, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_max_autotune, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_max_autotune_with_captured, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_mixed_dtypes_fails, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_multiple_score_mod_calls, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_multiple_score_mod_calls2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_njt_causal_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_njt_causal_float32, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_padded_dense_causal_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_recompile_changed_score_mod_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_seq_masking_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_silu_on_score_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_skip_odd_keys_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_skip_odd_keys_float32, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s0_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s0_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s0_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s0_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s1_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s1_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s1_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s1_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s2_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s2_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s2_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s2_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s3_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s3_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s3_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s0_v_s3_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s0_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s0_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s0_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s0_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s1_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s1_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s1_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s1_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s2_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s2_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s2_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s2_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s3_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s3_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s3_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s1_v_s3_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s0_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s0_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s0_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s0_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s1_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s1_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s1_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s1_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s2_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s2_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s2_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s2_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s3_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s3_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s3_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s2_v_s3_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s0_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s0_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s0_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s0_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s1_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s1_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s1_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s1_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s2_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s2_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s2_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s2_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s3_head_dims0, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s3_head_dims1, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s3_head_dims2, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_strided_inputs_float16_k_s3_v_s3_head_dims3, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_subgraph_respect_decompostion_float16, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_windowed_full_mask_vs_sdpa, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_windowed_no_mask_vs_sdpa, test/inductor/test_flex_decoding.py::TestFlexDecoding::test_windowed_partial_block_vs_sdpa 2024-08-06T21:58:01.7265709Z 2024-08-06T21:58:05.4384587Z Running dynamo/test_skip_non_tensor 1/1 ... [2024-08-06 21:58:05.437938] 2024-08-06T21:58:05.4388419Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'dynamo/test_skip_non_tensor.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:58:05.438421] 2024-08-06T21:58:10.0622337Z 2024-08-06T21:58:10.0623765Z dynamo/test_skip_non_tensor 1/1 was successful, full logs can be found in artifacts with path test/test-reports/dynamo.test_skip_non_tensor_1.1_25dff3f0ff7c9d6e_.log 2024-08-06T21:58:10.0627631Z Running 8 items in this shard: test/dynamo/test_skip_non_tensor.py::SkipNonTensorTests::test_add_skip, test/dynamo/test_skip_non_tensor.py::SkipNonTensorTests::test_add_tensor1, test/dynamo/test_skip_non_tensor.py::SkipNonTensorTests::test_add_tensor2, test/dynamo/test_skip_non_tensor.py::SkipNonTensorTests::test_add_tensor_dict, test/dynamo/test_skip_non_tensor.py::SkipNonTensorTests::test_add_tensor_list, test/dynamo/test_skip_non_tensor.py::SkipNonTensorTests::test_custom_list, test/dynamo/test_skip_non_tensor.py::SkipNonTensorTests::test_do_not_skip_side_effects, test/dynamo/test_skip_non_tensor.py::SkipNonTensorTests::test_recursive_list 2024-08-06T21:58:10.0630739Z 2024-08-06T21:58:13.7523581Z Running inductor/test_move_constructors_to_cuda 1/1 ... [2024-08-06 21:58:13.751873] 2024-08-06T21:58:13.7528363Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_move_constructors_to_cuda.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:58:13.752409] 2024-08-06T21:58:17.5920515Z 2024-08-06T21:58:17.5922202Z inductor/test_move_constructors_to_cuda 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_move_constructors_to_cuda_1.1_6012d3b3419e36c4_.log 2024-08-06T21:58:17.5923300Z 2024-08-06T21:58:21.4547224Z Running test_matmul_cuda 1/1 ... [2024-08-06 21:58:21.454148] 2024-08-06T21:58:21.4550750Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_matmul_cuda.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:58:21.454664] 2024-08-06T21:59:13.6845763Z 2024-08-06T21:59:13.6847242Z test_matmul_cuda 1/1 was successful, full logs can be found in artifacts with path test/test-reports/test_matmul_cuda_1.1_74baeb6da5259ad7_.log 2024-08-06T21:59:13.6863935Z Running 37 items in this shard: test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_alignment_cuda_float16, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_reduced_precision_size_10000_cuda_bfloat16, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_reduced_precision_size_10000_cuda_float16, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_reduced_precision_size_1000_cuda_bfloat16, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_reduced_precision_size_1000_cuda_float16, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_reduced_precision_size_100_cuda_bfloat16, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_reduced_precision_size_100_cuda_float16, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_size_10000_cuda_bfloat16, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_size_10000_cuda_float16, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_size_10000_cuda_float32, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_size_1000_cuda_bfloat16, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_size_1000_cuda_float16, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_size_1000_cuda_float32, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_size_100_cuda_bfloat16, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_size_100_cuda_float16, test/test_matmul_cuda.py::TestMatmulCudaCUDA::test_cublas_addmm_size_100_cuda_float32, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_error_message_fp8_pre_sm89_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_float32_output_errors_with_bias_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_float8_basics_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_float8_bias_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_float8_bias_relu_edgecase_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_float8_error_messages_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_float8_rowwise_scaling_sanity_use_fast_accum_False_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_float8_rowwise_scaling_sanity_use_fast_accum_True_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_float8_scale_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_float8_scale_fast_accum_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_non_divisible_leading_dim_bias_False_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_non_divisible_leading_dim_bias_True_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_scaled_mm_change_stride_bfloat16_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_scaled_mm_change_stride_float16_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_scaled_mm_change_stride_float32_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_scaled_mm_vs_emulated_bfloat16_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_scaled_mm_vs_emulated_float16_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_scaled_mm_vs_emulated_float32_cuda, test/test_matmul_cuda.py::TestFP8MatmulCudaCUDA::test_scaled_mm_vs_emulated_row_wise_bfloat16_cuda, test/test_matmul_cuda.py::TestMixedDtypesLinearCudaCUDA::test_mixed_dtypes_linear_cuda_bfloat16, test/test_matmul_cuda.py::TestMixedDtypesLinearCudaCUDA::test_mixed_dtypes_linear_cuda_float16 2024-08-06T21:59:13.6879904Z 2024-08-06T21:59:17.4243074Z Running dynamo/test_sdpa 1/1 ... [2024-08-06 21:59:17.423792] 2024-08-06T21:59:17.4247179Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'dynamo/test_sdpa.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:59:17.424275] 2024-08-06T21:59:21.9475672Z 2024-08-06T21:59:21.9476913Z dynamo/test_sdpa 1/1 was successful, full logs can be found in artifacts with path test/test-reports/dynamo.test_sdpa_1.1_bc8e6c4d8a4e0ef2_.log 2024-08-06T21:59:21.9479000Z Running 4 items in this shard: test/dynamo/test_sdpa.py::TestSDPA::test_graph_break_SDPAParams, test/dynamo/test_sdpa.py::TestSDPA::test_input_SDPAParams, test/dynamo/test_sdpa.py::TestSDPA::test_intermediate_attr_access_SDPAParams, test/dynamo/test_sdpa.py::TestSDPA::test_returns_SDPAParams 2024-08-06T21:59:21.9480460Z 2024-08-06T21:59:25.6940050Z Running inductor/test_metrics 1/1 ... [2024-08-06 21:59:25.693471] 2024-08-06T21:59:25.6943923Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_metrics.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:59:25.693977] 2024-08-06T21:59:29.6162640Z 2024-08-06T21:59:29.6164010Z inductor/test_metrics 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_metrics_1.1_a9ecdd0dd9f27f4e_.log 2024-08-06T21:59:29.6165063Z 2024-08-06T21:59:33.3953716Z Running inductor/test_debug_trace 1/1 ... [2024-08-06 21:59:33.394810] 2024-08-06T21:59:33.3957383Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_debug_trace.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:59:33.395262] 2024-08-06T21:59:45.8341187Z 2024-08-06T21:59:45.8342538Z inductor/test_debug_trace 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_debug_trace_1.1_8ab792b616f63b4e_.log 2024-08-06T21:59:45.8344953Z Running 2 items in this shard: test/inductor/test_debug_trace.py::TestDebugTrace::test_debug_multi_tempalte, test/inductor/test_debug_trace.py::TestDebugTrace::test_debug_trace 2024-08-06T21:59:45.8345873Z 2024-08-06T21:59:49.5897072Z Running test_modules 1/2 ... [2024-08-06 21:59:49.589213] 2024-08-06T21:59:49.5900927Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'test_modules.py', '-m', 'not serial', '--shard-id=1', '--num-shards=2', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 21:59:49.589700] 2024-08-06T22:01:05.2919349Z 2024-08-06T22:01:05.2923776Z inductor/test_torchinductor_dynamic_shapes 1/4 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_torchinductor_dynamic_shapes_1.4_cb3586c9672e681e_.log 2024-08-06T22:01:05.3024387Z Running 172 items in this shard: test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test__unsafe_masked_index_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_adaptive_avg_pool1d_argmax_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_adaptive_avg_pool2d_low_prec_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_adaptive_avg_pool_with_output_size_0_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_add_complex4_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_add_complex_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_add_const_float_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_add_const_int_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_addmm_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_angle_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_any_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_arange5_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_arange6_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_argmax_argmin1_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_argmax_argmin_with_nan_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_avg_pool2d_backward2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_avg_pool3d_backward4_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_batch_norm_2d_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_bitwise2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_bmm2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_buffer_copied_in_graph_with_different_shapes_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_builtins_round_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_builtins_round_int_ndigits_pos_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_builtins_round_int_ndigits_zero_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_cat_empty_index_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_cat_negative_dim_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_complex_memory_overlap_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_constant_pad_fill_dtype_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_constant_pad_float64_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_conv3d_channels_last_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_conv_bn_fuse_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_conv_with_as_strided_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_cudnn_rnn_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_custom_scan_op_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_dense_mask_index_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_diagonal_copy_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_dist_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_div1_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_div7_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_div9_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_div_zero_dim_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_dtypeview_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_exp_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_fallback_mutable_op_basic_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_fallback_mutable_op_list_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_float32_to_int32_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_float_index_expression_type_promotion_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_fmin_fmax_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_fmod_zero_dim_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_fractional_max_pool2d1_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_fractional_max_pool2d2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_full_like_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_gather2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_gelu_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_hardsigmoid_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_horizonal_fusion2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_index2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_index_propagation_device_assert_masked_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_index_propagation_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_index_propagation_flip_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_index_propagation_nested_indirect_indexing_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_index_put_deterministic_fallback_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_index_put_index_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_index_select_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_index_tensor_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_input_mutation4_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_int_input_dynamic_shapes_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_invalid_operand_issue1_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_issue102546_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_large_grid_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_large_offset_pointwise_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_layer_norm_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_leaky_relu_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_linear_float64_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_linspace1_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_linspace2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_log_softmax_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_logsumexp_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_low_memory_max_pool_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_masked_scatter_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_max_pool2d1_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_max_pool2d4_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_max_pool2d_with_indices_backward2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_max_pool2d_with_indices_backward3_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_max_pool2d_with_indices_backward4_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_max_pool2d_with_indices_backward6_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_min_max_reduction_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_misaligned_address_issue1_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_mixed_mm2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_mixed_mm3_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_mm_mixed_dtype_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_multi_gpu_device_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_neg_index_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_nll_loss_forward_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pattern_matcher_multi_user_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_permute2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_bessel_j0_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_erfc_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_erfcx_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_erfinv_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_gammainc_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_hermite_polynomial_h_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_hermite_polynomial_he_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_laguerre_polynomial_l_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_legendre_polynomial_p_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_log1p_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_log_ndtr_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_logit_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_modified_bessel_i0_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_modified_bessel_i1_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_multigammaln_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_ndtri_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pointwise_sinc_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_polar_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_pow_int_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_rand_like_deterministic_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_randn_like_empty_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_randn_with_dtype_and_device_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_reduction4_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_remainder_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_remove_no_ops_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_repeat_interleave_2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_resize_as_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_reuse_buffers_with_aliasing_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_round_correctness_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_round_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_scalar_output_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_scaled_dot_product_attention_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_scaled_dot_product_efficient_attention_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_scatter_add2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_scatter_reduce1_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_scatter_reduce3_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_should_pad_bench_for_bmm_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_sign_dtype_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_sin_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_slice4_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_slice_scatter3_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_slice_scatter5_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_slice_scatter_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_squeeze_varargs_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_stack_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_std_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_sum1_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_tanh_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_tensor2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_transpose_add_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_unbacked_floordiv_simplify_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_unsqueeze_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_unsqueeze_inplace_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_upsample_bilinear2d_a_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_upsample_cat_conv_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_upsample_nearest2d_backward_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_vdd_clamp_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_vectorized_ops_masked_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_view_uint8_through_differing_bitwidths_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_views2_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_xblock_divides_xnumel_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::DynamicShapesCpuTests::test_zero_dim_reductions_dynamic_shapes_cpu, test/inductor/test_torchinductor_dynamic_shapes.py::TestInductorDynamicCUDA::test_arithmetic_constant_folding_cuda, test/inductor/test_torchinductor_dynamic_shapes.py::TestInductorDynamicCUDA::test_dynamic_stride_nobreak_cuda, test/inductor/test_torchinductor_dynamic_shapes.py::TestInductorDynamicCUDA::test_float_is_integer_cuda, test/inductor/test_torchinductor_dynamic_shapes.py::TestInductorDynamicCUDA::test_float_item_neginf_cuda, test/inductor/test_torchinductor_dynamic_shapes.py::TestInductorDynamicCUDA::test_floor_cuda, test/inductor/test_torchinductor_dynamic_shapes.py::TestInductorDynamicCUDA::test_item_bool_nobreak_cuda, test/inductor/test_torchinductor_dynamic_shapes.py::TestInductorDynamicCUDA::test_item_to_inputs_kernel_nobreak_cuda, test/inductor/test_torchinductor_dynamic_shapes.py::TestInductorDynamicCUDA::test_math_ops_op3_cuda, test/inductor/test_torchinductor_dynamic_shapes.py::TestInductorDynamicCUDA::test_nonzero_no_realloc_cuda, test/inductor/test_torchinductor_dynamic_shapes.py::TestInductorDynamicCUDA::test_pad_dynamic_cuda, test/inductor/test_torchinductor_dynamic_shapes.py::TestInductorDynamicCUDA::test_sort_dynamic_shape_with_check_cuda, test/inductor/test_torchinductor_dynamic_shapes.py::TestInductorDynamicCUDA::test_unbacked_cat_backwards_save_data_dependent_cuda, test/inductor/test_torchinductor_dynamic_shapes.py::TestInductorDynamicCUDA::test_unbacked_matmul_cuda, test/inductor/test_torchinductor_dynamic_shapes.py::TestInductorDynamicCUDA::test_unbacked_reduction_cuda 2024-08-06T22:01:05.3121986Z 2024-08-06T22:01:09.0720317Z Running inductor/test_cuda_cpp_wrapper 1/1 ... [2024-08-06 22:01:09.071502] 2024-08-06T22:01:09.0723727Z Executing ['/opt/conda/envs/py_3.10/bin/python', '-bb', 'inductor/test_cuda_cpp_wrapper.py', '-m', 'not serial', '--shard-id=1', '--num-shards=1', '-v', '-vv', '-rfEX', '-p', 'no:xdist', '--use-pytest', '-x', '--reruns=2', '--import-slow-tests', '--import-disabled-tests'] ... [2024-08-06 22:01:09.071952] 2024-08-06T22:01:16.7234845Z 2024-08-06T22:01:16.7236222Z inductor/test_cuda_cpp_wrapper 1/1 was successful, full logs can be found in artifacts with path test/test-reports/inductor.test_cuda_cpp_wrapper_1.1_937d11bbc077ebd2_.log 2024-08-06T22:01:16.7237111Z 2024-08-06T22:09:06.7536575Z 2024-08-06T22:09:06.7537562Z test_modules 1/2 was successful, full logs can be found in artifacts with path test/test-reports/test_modules_1.2_8c1ccc7699308450_.log 2024-08-06T22:09:06.8311158Z Running 1796 items in this shard: test/test_modules.py::TestModuleCUDA::test_check_inplace_nn_CELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_check_inplace_nn_Mish_cuda_float64, test/test_modules.py::TestModuleCUDA::test_check_inplace_nn_ReLU6_cuda_float32, test/test_modules.py::TestModuleCUDA::test_check_inplace_nn_ReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_check_inplace_nn_SELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_check_inplace_nn_SELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_check_inplace_nn_SiLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_check_inplace_nn_SiLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_check_inplace_nn_Threshold_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_AdaptiveAvgPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_AdaptiveMaxPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_AdaptiveMaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_AdaptiveMaxPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_AdaptiveMaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_AdaptiveMaxPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_AdaptiveMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_AvgPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_AvgPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_AvgPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_AvgPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_AvgPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_BCELoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_BCEWithLogitsLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_BatchNorm1d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_BatchNorm2d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_BatchNorm2d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_BatchNorm3d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_BatchNorm3d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_BatchNorm3d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Bilinear_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Bilinear_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_CELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_CELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_CTCLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_CircularPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_CircularPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ConstantPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ConstantPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ConstantPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Conv1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ConvTranspose1d_cuda_complex128, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ConvTranspose1d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ConvTranspose1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ConvTranspose1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ConvTranspose2d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ConvTranspose2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ConvTranspose2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ConvTranspose3d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ConvTranspose3d_cuda_complex64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ConvTranspose3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_CosineEmbeddingLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_CrossEntropyLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Embedding_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_FractionalMaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_FractionalMaxPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_FractionalMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_GLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_GLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_GRUCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_GRU_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_GaussianNLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_GaussianNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_GroupNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_GroupNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Hardshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Hardswish_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Hardswish_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_HingeEmbeddingLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_HuberLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_InstanceNorm1d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_InstanceNorm2d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_InstanceNorm2d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_InstanceNorm3d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_KLDivLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_KLDivLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_L1Loss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_LPPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_LPPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_LPPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_LSTMCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_LSTM_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_LSTM_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_LazyConv1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_LazyConv1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_LazyConv3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_LazyConv3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_LazyConvTranspose1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Linear_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Linear_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_LogSigmoid_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_MaxPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_MaxPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_MaxPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Mish_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_MultiLabelMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_MultiLabelSoftMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_MultiLabelSoftMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_MultiMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_MultiMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_MultiheadAttention_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_MultiheadAttention_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_NLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_PReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_PoissonNLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_RMSNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_RNNCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ReflectionPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ReflectionPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ReplicationPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ReplicationPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ReplicationPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_SiLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Sigmoid_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_SmoothL1Loss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_SmoothL1Loss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_SoftMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Softmax_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Softplus_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Softshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Softshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Softsign_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Softsign_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Tanh_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Tanhshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_Tanhshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_TransformerDecoderLayer_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_TransformerEncoderLayer_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_TransformerEncoderLayer_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_TransformerEncoder_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_TransformerEncoder_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ZeroPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ZeroPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_cpu_gpu_parity_nn_ZeroPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_AdaptiveAvgPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_AdaptiveAvgPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_AdaptiveAvgPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_AdaptiveMaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_AdaptiveMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_BCELoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_BCELoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_BatchNorm1d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_BatchNorm1d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_BatchNorm1d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_BatchNorm2d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_BatchNorm3d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_BatchNorm3d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_BatchNorm3d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Bilinear_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_CELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_CTCLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_CircularPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_CircularPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_CircularPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_CircularPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_CircularPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ConstantPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ConstantPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ConstantPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ConstantPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Conv3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Conv3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ConvTranspose1d_cuda_complex128, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ConvTranspose2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ConvTranspose2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ConvTranspose3d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ConvTranspose3d_cuda_complex64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ConvTranspose3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_CrossEntropyLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Embedding_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Embedding_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_FractionalMaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_FractionalMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_GLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_GRUCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_GRU_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_GRU_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_GRU_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_GaussianNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_GroupNorm_cuda_bfloat16, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_GroupNorm_cuda_float16, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_GroupNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Hardswish_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Hardtanh_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_HingeEmbeddingLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_HuberLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_InstanceNorm1d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_InstanceNorm1d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_InstanceNorm1d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_InstanceNorm2d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_InstanceNorm2d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_InstanceNorm3d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_InstanceNorm3d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_KLDivLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_KLDivLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_L1Loss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_LPPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_LSTMCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_LSTM_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_LSTM_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_LayerNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_LazyConv1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_LazyConvTranspose1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_LazyConvTranspose1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_LazyConvTranspose2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_LazyConvTranspose2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Linear_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_LocalResponseNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_LogSigmoid_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_LogSoftmax_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_MSELoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_MSELoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_MarginRankingLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_MaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_MaxPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_MaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_MultiLabelMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_MultiLabelMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_MultiLabelSoftMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_MultiheadAttention_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_MultiheadAttention_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_MultiheadAttention_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_NLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_PReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_PReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_PoissonNLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_PoissonNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_RMSNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_RNNCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_RNN_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_RNN_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ReLU6_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ReflectionPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ReflectionPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ReflectionPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ReplicationPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ReplicationPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_SELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_SiLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Sigmoid_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_SmoothL1Loss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_SmoothL1Loss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_SoftMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Softmax2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Softmax_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Softmin_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Softplus_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Softshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Softshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Softsign_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Softsign_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Tanh_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Tanh_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Tanhshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_TransformerDecoderLayer_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_TransformerEncoderLayer_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_TransformerEncoderLayer_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_TransformerEncoderLayer_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_TransformerEncoder_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_TransformerEncoder_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_TransformerEncoder_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Transformer_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_Transformer_cuda_float64, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ZeroPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_device_ctx_init_nn_ZeroPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_errors_nn_CircularPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_errors_nn_CircularPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_errors_nn_CircularPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_errors_nn_GRUCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_errors_nn_GRU_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_errors_nn_LSTMCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_errors_nn_LSTMCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_errors_nn_LSTM_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_errors_nn_LSTM_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_errors_nn_RNNCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_errors_nn_RNNCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_errors_nn_RNN_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_errors_nn_RNN_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_AdaptiveMaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_AdaptiveMaxPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_AdaptiveMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_AvgPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_AvgPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_BCELoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_BCEWithLogitsLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_BCEWithLogitsLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_BatchNorm2d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_BatchNorm2d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_BatchNorm3d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_BatchNorm3d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_BatchNorm3d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Bilinear_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_CircularPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_CircularPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ConstantPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ConstantPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Conv1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Conv2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ConvTranspose2d_cuda_complex64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ConvTranspose3d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ConvTranspose3d_cuda_complex64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_CrossEntropyLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Embedding_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Embedding_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_FractionalMaxPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_FractionalMaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_FractionalMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_GELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_GLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_GLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_GRUCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_GRUCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_GRU_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_GaussianNLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_GaussianNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_GroupNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Hardshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Hardshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Hardswish_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_HingeEmbeddingLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_HuberLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_InstanceNorm1d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_InstanceNorm1d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_InstanceNorm2d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_InstanceNorm2d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_InstanceNorm3d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_KLDivLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_L1Loss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LPPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LPPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LSTMCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LSTMCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LSTM_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LSTM_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LazyConv2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LazyConv3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LazyConvTranspose1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LazyConvTranspose2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LazyConvTranspose3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LazyConvTranspose3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Linear_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LocalResponseNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LocalResponseNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LogSigmoid_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_LogSoftmax_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_MSELoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_MSELoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_MarginRankingLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_MaxPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_MaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Mish_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_MultiLabelMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_MultiheadAttention_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_MultiheadAttention_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_NLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_NLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_PReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_PoissonNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_RMSNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_RMSNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_RNNCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_RNN_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_RNN_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ReLU6_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ReLU6_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ReflectionPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ReflectionPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ReflectionPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ReflectionPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ReplicationPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ReplicationPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ReplicationPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ReplicationPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_SELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_SiLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_SoftMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Softmax2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Softmax_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Softmin_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Softmin_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Softplus_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Softplus_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Softshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Softshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Softsign_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Tanh_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Tanh_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Tanhshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Threshold_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Threshold_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_TransformerDecoderLayer_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_TransformerDecoderLayer_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_TransformerEncoderLayer_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_TransformerEncoderLayer_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_TransformerEncoder_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_TransformerEncoder_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_TransformerEncoder_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Transformer_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_Transformer_cuda_float64, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ZeroPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_factory_kwargs_nn_ZeroPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_AdaptiveAvgPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_AdaptiveAvgPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_AdaptiveAvgPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_AdaptiveMaxPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_AdaptiveMaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_AdaptiveMaxPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_AdaptiveMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_AvgPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_AvgPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_AvgPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_AvgPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_AvgPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_BCELoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_BCEWithLogitsLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_BCEWithLogitsLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_BatchNorm1d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_BatchNorm1d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_BatchNorm2d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_BatchNorm2d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_BatchNorm3d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_BatchNorm3d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_BatchNorm3d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_Bilinear_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_CELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_CELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_CTCLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_CircularPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_ConstantPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_ConstantPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_Conv1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_Conv2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_Conv3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_ConvTranspose1d_cuda_complex64, test/test_modules.py::TestModuleCUDA::test_forward_nn_ConvTranspose1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_ConvTranspose2d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_forward_nn_ConvTranspose2d_cuda_complex64, test/test_modules.py::TestModuleCUDA::test_forward_nn_ConvTranspose2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_ConvTranspose2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_ConvTranspose3d_cuda_complex128, test/test_modules.py::TestModuleCUDA::test_forward_nn_ConvTranspose3d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_forward_nn_ConvTranspose3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_CrossEntropyLoss_cuda_float16, test/test_modules.py::TestModuleCUDA::test_forward_nn_CrossEntropyLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_CrossEntropyLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_ELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_FractionalMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_GLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_GRUCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_GRU_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_GroupNorm_cuda_bfloat16, test/test_modules.py::TestModuleCUDA::test_forward_nn_GroupNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_Hardshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_Hardshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_Hardswish_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_Hardtanh_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_Hardtanh_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_HuberLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_HuberLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_InstanceNorm1d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_InstanceNorm2d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_InstanceNorm2d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_InstanceNorm3d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_InstanceNorm3d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_L1Loss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_LPPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_LPPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_LPPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_LSTM_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_LayerNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_LazyConv1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_LazyConv2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_LazyConv3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_LazyConvTranspose1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_LazyConvTranspose3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_LazyConvTranspose3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_LeakyReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_LocalResponseNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_LogSoftmax_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_LogSoftmax_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_MarginRankingLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_MaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_MaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_MaxPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_MaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_MultiLabelSoftMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_MultiheadAttention_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_MultiheadAttention_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_NLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_PReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_RMSNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_RMSNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_RNNCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_RNNCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_RNN_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_RNN_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_RNN_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_RNN_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_ReLU6_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_ReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_ReflectionPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_ReplicationPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_ReplicationPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_SELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_SELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_SiLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_SiLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_Sigmoid_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_SmoothL1Loss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_SoftMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_SoftMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_Softmax2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_Softmax_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_Softmin_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_Softmin_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_Softshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_Softsign_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_Tanh_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_Tanh_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_TransformerDecoderLayer_cuda_float64, test/test_modules.py::TestModuleCUDA::test_forward_nn_TransformerEncoder_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_TransformerEncoder_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_forward_nn_ZeroPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_AdaptiveAvgPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_AdaptiveAvgPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_AdaptiveMaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_AdaptiveMaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_AvgPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_AvgPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_Bilinear_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_CircularPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_ConstantPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_ConstantPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_Conv1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_Conv3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_ConvTranspose2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_ConvTranspose3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_CosineEmbeddingLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_CrossEntropyLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_ELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_GLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_GRU_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_GaussianNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_Hardshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_Hardswish_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_Hardtanh_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_HingeEmbeddingLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_HuberLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_InstanceNorm1d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_InstanceNorm1d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_InstanceNorm2d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_InstanceNorm3d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_LSTMCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_LSTM_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_LazyConv2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_LazyConvTranspose1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_LazyConvTranspose3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_LeakyReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_Linear_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_LocalResponseNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_LogSigmoid_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_MSELoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_MarginRankingLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_MaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_MultiLabelMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_MultiLabelSoftMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_MultiheadAttention_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_NLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_PoissonNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_RMSNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_RNN_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_ReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_ReflectionPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_ReflectionPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_ReflectionPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_ReplicationPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_ReplicationPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_SELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_SiLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_Softmax2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_Softmin_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_Softshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_Softsign_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_Threshold_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_TransformerDecoderLayer_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_TransformerEncoderLayer_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_grad_nn_TransformerEncoderLayer_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_AdaptiveAvgPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_AdaptiveMaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_AdaptiveMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_AvgPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_AvgPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_BatchNorm1d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_BatchNorm2d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_BatchNorm3d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_BatchNorm3d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_Bilinear_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_CELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_CTCLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_CircularPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_ConstantPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_ConstantPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_Conv1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_Conv2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_Conv3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_ConvTranspose1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_ConvTranspose3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_CrossEntropyLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_GELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_GLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_GRUCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_GRU_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_GRU_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_GroupNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_Hardtanh_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_HuberLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_InstanceNorm1d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_InstanceNorm1d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_InstanceNorm3d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_InstanceNorm3d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_L1Loss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_LPPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_LSTMCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_LSTM_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_LazyConvTranspose1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_LazyConvTranspose2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_LazyConvTranspose3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_LeakyReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_LocalResponseNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_LogSigmoid_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_MarginRankingLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_MaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_MaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_MaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_Mish_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_MultiLabelMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_MultiMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_NLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_PReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_PoissonNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_RMSNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_RNNCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_ReflectionPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_ReflectionPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_ReflectionPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_ReplicationPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_ReplicationPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_SiLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_SoftMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_Softmax2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_Softplus_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_Softshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_Tanhshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_Threshold_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_TransformerEncoderLayer_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_TransformerEncoder_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_TransformerEncoder_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_Transformer_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_ZeroPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_ZeroPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_gradgrad_nn_ZeroPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_AdaptiveAvgPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_AdaptiveAvgPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_AdaptiveAvgPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_AdaptiveAvgPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_AdaptiveMaxPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_AdaptiveMaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_AdaptiveMaxPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_AdaptiveMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_AvgPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_AvgPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_AvgPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_AvgPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_BCELoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_BCEWithLogitsLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_BatchNorm1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_BatchNorm2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_BatchNorm3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Bilinear_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_CELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_CELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_CTCLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_CTCLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_CircularPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_CircularPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ConstantPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ConstantPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ConstantPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Conv1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Conv2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Conv3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ConvTranspose1d_cuda_complex64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ConvTranspose1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ConvTranspose2d_cuda_complex128, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ConvTranspose2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ConvTranspose2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ConvTranspose3d_cuda_complex128, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ConvTranspose3d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ConvTranspose3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_CosineEmbeddingLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_CrossEntropyLoss_cuda_float16, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_CrossEntropyLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_CrossEntropyLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Embedding_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_FractionalMaxPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_FractionalMaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_FractionalMaxPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_FractionalMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_GLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_GRUCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_GroupNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Hardshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Hardshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Hardswish_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Hardswish_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Hardtanh_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Hardtanh_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_HuberLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_InstanceNorm2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_InstanceNorm3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_KLDivLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_KLDivLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LPPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LPPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LPPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LSTMCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LSTM_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LayerNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LayerNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LazyConv1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LazyConv1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LazyConv2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LazyConv2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LazyConvTranspose1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LazyConvTranspose2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LeakyReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LeakyReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Linear_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LocalResponseNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LocalResponseNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LogSigmoid_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LogSigmoid_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_LogSoftmax_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_MSELoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_MSELoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_MarginRankingLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_MaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_MaxPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_MaxPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_MaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Mish_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_MultiLabelSoftMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_MultiheadAttention_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_NLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_PReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_PReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_PoissonNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_RMSNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_RNNCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ReflectionPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ReflectionPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ReflectionPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ReflectionPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ReplicationPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ReplicationPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_SELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_SELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_SiLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_SmoothL1Loss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_SoftMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Softmax2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Softmin_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Softmin_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Softshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Tanhshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Tanhshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Threshold_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_TransformerDecoderLayer_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_TransformerDecoderLayer_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_TransformerEncoder_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Transformer_cuda_float32, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_Transformer_cuda_float64, test/test_modules.py::TestModuleCUDA::test_if_train_and_eval_modes_differ_nn_ZeroPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_AdaptiveAvgPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_AdaptiveAvgPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_AdaptiveAvgPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_AdaptiveMaxPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_AdaptiveMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_AvgPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_AvgPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_BCEWithLogitsLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_BCEWithLogitsLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_BatchNorm1d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_BatchNorm1d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_BatchNorm1d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_BatchNorm1d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_BatchNorm2d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_BatchNorm2d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Bilinear_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_CircularPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ConstantPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ConstantPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Conv1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Conv3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ConvTranspose1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ConvTranspose1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ConvTranspose2d_cuda_complex128, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ConvTranspose2d_cuda_complex64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ConvTranspose3d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ConvTranspose3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_CosineEmbeddingLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_CosineEmbeddingLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_CrossEntropyLoss_cuda_float16, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_CrossEntropyLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_CrossEntropyLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Embedding_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Embedding_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_FractionalMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_GLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_GLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_GRU_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_GRU_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_GRU_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_GaussianNLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_GaussianNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_GroupNorm_cuda_bfloat16, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_GroupNorm_cuda_float16, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_GroupNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Hardshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_HingeEmbeddingLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_InstanceNorm1d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_InstanceNorm1d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_InstanceNorm2d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_InstanceNorm2d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_InstanceNorm3d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_InstanceNorm3d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_KLDivLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_L1Loss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LPPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LPPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LPPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LSTMCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LSTMCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LSTM_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LSTM_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LSTM_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LayerNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LayerNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LazyConv1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LazyConv3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LazyConv3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LazyConvTranspose1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LazyConvTranspose1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LazyConvTranspose2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Linear_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Linear_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LocalResponseNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_LogSoftmax_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_MSELoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_MarginRankingLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_MaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_MultiLabelMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_MultiLabelSoftMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_MultiLabelSoftMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_MultiheadAttention_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_NLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_PoissonNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_RMSNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_RNNCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_RNN_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_RNN_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_RNN_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ReflectionPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ReflectionPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ReflectionPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ReflectionPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ReflectionPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ReplicationPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ReplicationPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ReplicationPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_SELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_SiLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Sigmoid_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Sigmoid_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_SmoothL1Loss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_SmoothL1Loss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Softmax2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Softmax2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Softmax_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Softmax_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Softplus_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Softsign_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Softsign_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Tanh_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Tanhshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Threshold_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_Threshold_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_TransformerDecoderLayer_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_TransformerEncoderLayer_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_TransformerEncoder_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_TransformerEncoder_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_TransformerEncoder_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ZeroPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ZeroPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ZeroPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_memory_format_nn_ZeroPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_AdaptiveAvgPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_AdaptiveAvgPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_AdaptiveAvgPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_AdaptiveMaxPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_AvgPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_AvgPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_AvgPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_AvgPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_BCELoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_BCELoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_BCEWithLogitsLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_BatchNorm2d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_BatchNorm2d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_BatchNorm3d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_BatchNorm3d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Bilinear_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_CELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_CTCLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_CircularPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_CircularPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_CircularPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_CircularPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ConstantPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ConstantPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ConstantPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ConstantPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Conv2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Conv2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Conv3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ConvTranspose1d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ConvTranspose2d_cuda_complex128, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ConvTranspose2d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ConvTranspose2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ConvTranspose3d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ConvTranspose3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ConvTranspose3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_CosineEmbeddingLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_CrossEntropyLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_CrossEntropyLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_FractionalMaxPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_FractionalMaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_FractionalMaxPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_FractionalMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_GELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_GELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_GLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_GRU_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_GRU_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_GRU_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_GaussianNLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_GaussianNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Hardshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Hardswish_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Hardswish_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Hardtanh_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_HingeEmbeddingLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_HingeEmbeddingLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_InstanceNorm1d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_InstanceNorm1d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_InstanceNorm1d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_InstanceNorm2d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_InstanceNorm2d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_InstanceNorm2d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_InstanceNorm3d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_KLDivLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_L1Loss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_LPPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_LPPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_LPPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_LPPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_LSTMCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_LSTM_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_LSTM_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_LayerNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_LazyConv1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_LazyConv2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_LazyConvTranspose1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_LazyConvTranspose3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_LeakyReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Linear_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_LogSoftmax_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_MSELoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_MarginRankingLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_MaxPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_MaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_MaxPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_MaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_MaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Mish_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_MultiMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_NLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_NLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_PReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_PoissonNLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_PoissonNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_RMSNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_RNNCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_RNN_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_RNN_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_RNN_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ReLU6_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ReflectionPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ReplicationPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ReplicationPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ReplicationPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ReplicationPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_SELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_SiLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Sigmoid_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Sigmoid_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_SmoothL1Loss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_SmoothL1Loss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Softmin_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Softmin_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Softplus_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Softshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Softsign_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Tanh_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Tanh_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Tanhshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Threshold_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_TransformerDecoderLayer_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_TransformerEncoderLayer_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_TransformerEncoderLayer_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_TransformerEncoderLayer_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_TransformerEncoder_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_TransformerEncoder_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Transformer_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_Transformer_cuda_float64, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ZeroPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ZeroPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ZeroPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_multiple_device_transfer_nn_ZeroPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_AdaptiveAvgPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_AdaptiveAvgPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_AdaptiveMaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_AvgPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_AvgPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_AvgPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_BCELoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_BatchNorm1d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_BatchNorm1d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_BatchNorm1d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_BatchNorm2d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_BatchNorm2d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_BatchNorm3d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_BatchNorm3d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_BatchNorm3d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Bilinear_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Bilinear_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_CTCLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_CircularPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_CircularPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_CircularPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ConstantPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ConstantPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ConstantPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ConstantPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Conv1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Conv3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ConvTranspose1d_cuda_complex64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ConvTranspose2d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ConvTranspose2d_cuda_complex64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ConvTranspose2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ConvTranspose3d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ConvTranspose3d_cuda_complex64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_CrossEntropyLoss_cuda_float16, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_CrossEntropyLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Embedding_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_FractionalMaxPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_FractionalMaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_FractionalMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_GELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_GRUCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_GaussianNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_GroupNorm_cuda_bfloat16, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Hardshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Hardtanh_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_InstanceNorm1d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_InstanceNorm2d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_InstanceNorm2d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_InstanceNorm3d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_InstanceNorm3d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_InstanceNorm3d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LPPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LPPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LPPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LPPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LSTMCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LSTMCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LSTM_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LSTM_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LSTM_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LSTM_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LayerNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LayerNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LazyConv2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LazyConv2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LazyConv3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LazyConvTranspose1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LazyConvTranspose2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LazyConvTranspose2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LazyConvTranspose3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LeakyReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LocalResponseNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LocalResponseNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LogSigmoid_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_LogSoftmax_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_MarginRankingLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_MaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_MaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_MaxPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Mish_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_MultiLabelSoftMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_MultiheadAttention_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_NLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_PReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_PReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_PoissonNLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_RMSNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_RMSNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_RNNCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_RNNCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_RNN_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_RNN_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_RNN_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ReLU6_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ReflectionPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ReflectionPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ReflectionPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ReplicationPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ReplicationPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_SiLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Sigmoid_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_SmoothL1Loss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_SmoothL1Loss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_SoftMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Softmax2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Softmax2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Softmax_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Softmax_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Softmin_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Softmin_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Softplus_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Softplus_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Softshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Threshold_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Threshold_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_TransformerEncoderLayer_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_TransformerEncoderLayer_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_TransformerEncoder_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_TransformerEncoder_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Transformer_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_Transformer_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ZeroPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ZeroPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_non_contiguous_tensors_nn_ZeroPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_AdaptiveAvgPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_AdaptiveAvgPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_AdaptiveAvgPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_AdaptiveAvgPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_AdaptiveMaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_AdaptiveMaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_AvgPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_AvgPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_AvgPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_AvgPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_AvgPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_BCELoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_BCEWithLogitsLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_BCEWithLogitsLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_BatchNorm1d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_BatchNorm1d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_BatchNorm2d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_BatchNorm3d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_BatchNorm3d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_Bilinear_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_CTCLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_CircularPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_CircularPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_CircularPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_CircularPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_ConstantPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_ConstantPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_ConvTranspose1d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_repr_nn_ConvTranspose1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_ConvTranspose3d_cuda_complex128, test/test_modules.py::TestModuleCUDA::test_repr_nn_ConvTranspose3d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_repr_nn_ConvTranspose3d_cuda_complex64, test/test_modules.py::TestModuleCUDA::test_repr_nn_ConvTranspose3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_CrossEntropyLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_ELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_Embedding_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_FractionalMaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_GLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_GRUCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_GRU_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_GaussianNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_GroupNorm_cuda_bfloat16, test/test_modules.py::TestModuleCUDA::test_repr_nn_GroupNorm_cuda_float16, test/test_modules.py::TestModuleCUDA::test_repr_nn_GroupNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_Hardshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_Hardswish_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_Hardswish_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_HingeEmbeddingLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_HingeEmbeddingLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_HuberLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_InstanceNorm2d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_InstanceNorm2d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_InstanceNorm3d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_L1Loss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_LPPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_LPPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_LPPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_LPPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_LSTM_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_LSTM_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_LSTM_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_LayerNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_LayerNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_LazyConv1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_LazyConv2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_LazyConv3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_LazyConv3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_LazyConvTranspose1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_LazyConvTranspose2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_LazyConvTranspose3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_Linear_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_LogSigmoid_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_LogSigmoid_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_LogSoftmax_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_LogSoftmax_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_MSELoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_MaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_MaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_MaxPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_MaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_Mish_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_MultiLabelMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_MultiLabelSoftMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_MultiLabelSoftMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_MultiMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_MultiheadAttention_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_MultiheadAttention_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_PReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_PReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_PoissonNLLLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_RMSNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_RMSNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_RNNCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_ReLU6_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_ReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_ReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_ReflectionPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_ReflectionPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_ReflectionPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_ReflectionPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_ReplicationPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_ReplicationPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_ReplicationPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_SiLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_SmoothL1Loss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_SmoothL1Loss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_SoftMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_SoftMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_Softmax2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_Softmax_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_Softmin_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_Softplus_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_Softshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_Softshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_Softsign_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_TransformerEncoderLayer_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_TransformerEncoderLayer_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_TransformerEncoderLayer_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_TransformerEncoder_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_TransformerEncoder_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_TransformerEncoder_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_repr_nn_Transformer_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_ZeroPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_repr_nn_ZeroPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_AdaptiveAvgPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_AdaptiveAvgPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_AdaptiveAvgPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_AdaptiveMaxPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_AdaptiveMaxPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_AvgPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_AvgPool1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_AvgPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_BCELoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_BatchNorm2d_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_BatchNorm2d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_BatchNorm3d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_BatchNorm3d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_BatchNorm3d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Bilinear_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_CELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_CELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_CircularPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_CircularPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_CircularPad2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_CircularPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ConstantPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ConstantPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Conv1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Conv1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Conv2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Conv2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Conv3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ConvTranspose1d_cuda_complex128, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ConvTranspose1d_cuda_complex64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ConvTranspose1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ConvTranspose2d_cuda_complex32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ConvTranspose2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ConvTranspose3d_cuda_complex128, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ConvTranspose3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_CosineEmbeddingLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_CosineEmbeddingLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_CrossEntropyLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Embedding_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Embedding_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_FractionalMaxPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_GLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_GRUCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_GRU_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_GRU_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_GaussianNLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_GroupNorm_cuda_float16, test/test_modules.py::TestModuleCUDA::test_save_load_nn_GroupNorm_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Hardshrink_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Hardtanh_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_HingeEmbeddingLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_InstanceNorm1d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_InstanceNorm1d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_InstanceNorm1d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_InstanceNorm2d_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_InstanceNorm2d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_InstanceNorm3d_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_InstanceNorm3d_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_KLDivLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_L1Loss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_L1Loss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LPPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LPPool2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LPPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LPPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LPPool3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LSTMCell_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LSTM_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LayerNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LazyConv3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LazyConv3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LazyConvTranspose1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LazyConvTranspose1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LazyConvTranspose2d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LazyConvTranspose3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LeakyReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LeakyReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LogSigmoid_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_LogSoftmax_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_MSELoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_MarginRankingLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_MaxPool1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_MaxPool2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_MaxPool3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_MultiLabelMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_MultiLabelSoftMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_MultiMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_MultiheadAttention_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_MultiheadAttention_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_NLLLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_PReLU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_PReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_RMSNorm_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_RNNCell_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_RNN_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ReLU6_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ReLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ReflectionPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ReflectionPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ReflectionPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ReplicationPad1d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ReplicationPad1d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ReplicationPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_SELU_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_SELU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_SiLU_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_SmoothL1Loss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_SoftMarginLoss_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_SoftMarginLoss_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Softmax_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Softmin_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Softmin_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Softshrink_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Softsign_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Tanh_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_TransformerDecoderLayer_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_TransformerEncoderLayer_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_TransformerEncoderLayer_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_TransformerEncoderLayer_train_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_TransformerEncoder_eval_mode_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_TransformerEncoder_eval_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_TransformerEncoder_train_mode_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_Transformer_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ZeroPad2d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ZeroPad3d_cuda_float32, test/test_modules.py::TestModuleCUDA::test_save_load_nn_ZeroPad3d_cuda_float64, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_AdaptiveAvgPool1d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_AdaptiveAvgPool2d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_AdaptiveAvgPool3d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_AdaptiveAvgPool3d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_AdaptiveMaxPool1d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_AdaptiveMaxPool2d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_AdaptiveMaxPool3d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_AdaptiveMaxPool3d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_AvgPool1d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_AvgPool3d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_BatchNorm1d_eval_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_BatchNorm1d_eval_mode_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_BatchNorm1d_train_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_BatchNorm2d_eval_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_BatchNorm2d_eval_mode_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_BatchNorm2d_train_mode_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_BatchNorm3d_eval_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_BatchNorm3d_train_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Bilinear_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_CELU_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_CELU_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_CTCLoss_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_CircularPad2d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_CircularPad3d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ConstantPad2d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ConstantPad3d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ConstantPad3d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Conv1d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Conv2d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Conv2d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ConvTranspose1d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ConvTranspose2d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ConvTranspose2d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ConvTranspose3d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_CosineEmbeddingLoss_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_CrossEntropyLoss_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ELU_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_FractionalMaxPool2d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_FractionalMaxPool2d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_FractionalMaxPool3d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_FractionalMaxPool3d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_GELU_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_GELU_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_GLU_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_GLU_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_GRUCell_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_GRU_train_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_GroupNorm_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Hardshrink_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Hardshrink_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Hardswish_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Hardtanh_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_HingeEmbeddingLoss_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_InstanceNorm1d_eval_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_InstanceNorm1d_eval_mode_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_InstanceNorm1d_train_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_InstanceNorm2d_eval_mode_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_InstanceNorm3d_eval_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_InstanceNorm3d_train_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_InstanceNorm3d_train_mode_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_KLDivLoss_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_LPPool1d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_LPPool3d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_LPPool3d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_LSTMCell_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_LSTM_eval_mode_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_LSTM_train_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_LeakyReLU_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_LeakyReLU_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Linear_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_LocalResponseNorm_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_LogSigmoid_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_LogSoftmax_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_LogSoftmax_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_MSELoss_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_MSELoss_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_MarginRankingLoss_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_MaxPool1d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_MaxPool2d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_MaxPool2d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Mish_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_MultiLabelMarginLoss_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_MultiLabelMarginLoss_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_MultiMarginLoss_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_MultiheadAttention_eval_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_MultiheadAttention_train_mode_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_NLLLoss_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_NLLLoss_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_RMSNorm_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_RNN_eval_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_RNN_train_mode_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ReflectionPad1d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ReflectionPad2d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ReflectionPad3d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ReplicationPad1d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ReplicationPad1d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ReplicationPad2d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ReplicationPad3d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ReplicationPad3d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_SiLU_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Sigmoid_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Sigmoid_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_SmoothL1Loss_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_SmoothL1Loss_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_SoftMarginLoss_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Softmax2d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Softmax_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Softplus_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Softshrink_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Softsign_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Softsign_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Tanh_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Tanhshrink_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_Threshold_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_TransformerEncoderLayer_eval_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_TransformerEncoderLayer_eval_mode_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_TransformerEncoder_eval_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_TransformerEncoder_train_mode_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_TransformerEncoder_train_mode_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ZeroPad1d_swap_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ZeroPad1d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ZeroPad2d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_empty_nn_ZeroPad3d_swap_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveAvgPool1d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveAvgPool1d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveAvgPool2d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveAvgPool3d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveAvgPool3d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveAvgPool3d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveMaxPool1d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveMaxPool1d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveMaxPool1d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveMaxPool2d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveMaxPool2d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveMaxPool2d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveMaxPool3d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveMaxPool3d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AdaptiveMaxPool3d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AvgPool1d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AvgPool1d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AvgPool2d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AvgPool2d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AvgPool2d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AvgPool3d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AvgPool3d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_AvgPool3d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BCELoss_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BCEWithLogitsLoss_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BCEWithLogitsLoss_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BCEWithLogitsLoss_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BatchNorm1d_eval_mode_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BatchNorm1d_eval_mode_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BatchNorm1d_eval_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BatchNorm1d_eval_mode_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BatchNorm1d_train_mode_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BatchNorm1d_train_mode_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BatchNorm1d_train_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BatchNorm2d_eval_mode_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BatchNorm2d_eval_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BatchNorm2d_eval_mode_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BatchNorm2d_train_mode_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BatchNorm2d_train_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BatchNorm3d_eval_mode_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_BatchNorm3d_eval_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Bilinear_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Bilinear_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_CELU_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_CTCLoss_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_CTCLoss_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_CircularPad1d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_CircularPad2d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_CircularPad2d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_CircularPad2d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_CircularPad3d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ConstantPad1d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ConstantPad2d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Conv1d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Conv1d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Conv1d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Conv2d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Conv2d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Conv3d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Conv3d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Conv3d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ConvTranspose1d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ConvTranspose1d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ConvTranspose2d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ConvTranspose2d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ConvTranspose3d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ConvTranspose3d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ConvTranspose3d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ConvTranspose3d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_CosineEmbeddingLoss_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_CosineEmbeddingLoss_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_CosineEmbeddingLoss_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_CrossEntropyLoss_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_CrossEntropyLoss_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ELU_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ELU_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ELU_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ELU_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Embedding_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Embedding_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_FractionalMaxPool2d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_FractionalMaxPool3d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_FractionalMaxPool3d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_FractionalMaxPool3d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GELU_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GELU_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GLU_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GLU_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GLU_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GLU_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GRUCell_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GRUCell_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GRUCell_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GRU_eval_mode_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GRU_eval_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GRU_train_mode_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GRU_train_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GaussianNLLLoss_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GaussianNLLLoss_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GaussianNLLLoss_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GroupNorm_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GroupNorm_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_GroupNorm_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Hardshrink_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Hardshrink_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Hardshrink_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Hardshrink_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Hardswish_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Hardswish_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Hardtanh_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Hardtanh_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_HingeEmbeddingLoss_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_HingeEmbeddingLoss_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_HingeEmbeddingLoss_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_HuberLoss_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_HuberLoss_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_HuberLoss_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_InstanceNorm1d_train_mode_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_InstanceNorm1d_train_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_InstanceNorm1d_train_mode_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_InstanceNorm2d_eval_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_InstanceNorm2d_train_mode_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_InstanceNorm2d_train_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_InstanceNorm2d_train_mode_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_InstanceNorm3d_eval_mode_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_InstanceNorm3d_eval_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_InstanceNorm3d_eval_mode_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_InstanceNorm3d_train_mode_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_InstanceNorm3d_train_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_KLDivLoss_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_L1Loss_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_L1Loss_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LPPool1d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LPPool1d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LPPool1d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LPPool2d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LPPool2d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LPPool3d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LPPool3d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LSTMCell_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LSTM_eval_mode_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LSTM_eval_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LSTM_train_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LSTM_train_mode_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LayerNorm_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LeakyReLU_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Linear_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LocalResponseNorm_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LocalResponseNorm_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LogSigmoid_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LogSigmoid_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LogSoftmax_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_LogSoftmax_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MSELoss_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MSELoss_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MSELoss_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MarginRankingLoss_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MaxPool1d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MaxPool2d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MaxPool3d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Mish_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Mish_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Mish_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiLabelMarginLoss_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiLabelMarginLoss_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiLabelMarginLoss_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiLabelSoftMarginLoss_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiLabelSoftMarginLoss_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiLabelSoftMarginLoss_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiLabelSoftMarginLoss_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiMarginLoss_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiMarginLoss_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiheadAttention_eval_mode_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiheadAttention_eval_mode_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiheadAttention_eval_mode_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiheadAttention_train_mode_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiheadAttention_train_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_MultiheadAttention_train_mode_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_NLLLoss_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_PReLU_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_PReLU_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_PoissonNLLLoss_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_PoissonNLLLoss_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_RMSNorm_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_RMSNorm_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_RNNCell_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_RNNCell_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_RNN_eval_mode_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_RNN_train_mode_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_RNN_train_mode_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReLU6_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReLU_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReLU_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReflectionPad1d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReflectionPad1d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReflectionPad2d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReflectionPad2d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReflectionPad2d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReflectionPad3d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReflectionPad3d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReplicationPad1d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReplicationPad1d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReplicationPad2d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReplicationPad2d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReplicationPad3d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ReplicationPad3d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_SELU_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_SELU_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_SELU_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Sigmoid_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Sigmoid_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Sigmoid_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_SmoothL1Loss_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_SmoothL1Loss_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_SoftMarginLoss_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_SoftMarginLoss_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Softmax2d_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Softmax2d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Softmax_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Softmax_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Softmin_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Softmin_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Softplus_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Softshrink_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Softsign_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Softsign_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Tanh_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Tanhshrink_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Tanhshrink_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Threshold_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_TransformerDecoderLayer_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_TransformerDecoderLayer_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_TransformerDecoderLayer_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_TransformerEncoderLayer_eval_mode_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_TransformerEncoderLayer_train_mode_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_TransformerEncoderLayer_train_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_TransformerEncoderLayer_train_mode_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_TransformerEncoder_eval_mode_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_TransformerEncoder_eval_mode_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_TransformerEncoder_eval_mode_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_TransformerEncoder_train_mode_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Transformer_swap_False_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_Transformer_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ZeroPad1d_swap_False_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ZeroPad1d_swap_True_set_grad_True_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ZeroPad2d_swap_True_set_grad_False_cuda_float32, test/test_modules.py::TestModuleCUDA::test_to_nn_ZeroPad3d_swap_False_set_grad_True_cuda_float32 2024-08-06T22:09:06.9051726Z 2024-08-06T22:09:07.8137773Z 2024-08-06T22:09:07.8138797Z real 87m13.250s 2024-08-06T22:09:07.8139359Z user 152m51.203s 2024-08-06T22:09:07.8139658Z sys 13m23.423s 2024-08-06T22:09:07.8139974Z + assert_git_not_dirty 2024-08-06T22:09:07.8140391Z + [[ linux-focal-cuda12.1-py3.10-gcc9 != *rocm* ]] 2024-08-06T22:09:07.8140883Z + [[ linux-focal-cuda12.1-py3.10-gcc9 != *xla* ]] 2024-08-06T22:09:07.8144925Z ++ git status --porcelain 2024-08-06T22:09:07.8146074Z ++ grep -v '?? third_party' 2024-08-06T22:09:10.7824186Z ++ true 2024-08-06T22:09:10.7826020Z + git_status= 2024-08-06T22:09:10.7826384Z + [[ -n '' ]] 2024-08-06T22:09:10.7826750Z + test_aten 2024-08-06T22:09:10.7827362Z + echo 'Running ATen tests with pytorch lib' 2024-08-06T22:09:10.7827885Z Running ATen tests with pytorch lib 2024-08-06T22:09:10.7828296Z + [[ -n '' ]] 2024-08-06T22:09:10.7828861Z + echo 'Running test with the build folder' 2024-08-06T22:09:10.7829284Z Running test with the build folder 2024-08-06T22:09:10.7829680Z + TEST_BASE_DIR=build/bin 2024-08-06T22:09:10.7831099Z + ln -sf /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libc10.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libc10_cuda.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libc10d_cuda_test.so build/bin 2024-08-06T22:09:10.7844138Z + ln -sf /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libcaffe2_nvrtc.so build/bin 2024-08-06T22:09:10.7856126Z + ln -sf '/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libmkldnn*' build/bin 2024-08-06T22:09:10.7872621Z + ln -sf '/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libnccl*' build/bin 2024-08-06T22:09:10.7888752Z + ln -sf /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libtorch.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libtorch_cpu.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libtorch_cuda_linalg.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libtorch_global_deps.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libtorch_python.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libtorchbind_test.so build/bin 2024-08-06T22:09:10.7899248Z + ls build/bin 2024-08-06T22:09:10.7921264Z BackoffTest 2024-08-06T22:09:10.7921934Z CppSignature_test 2024-08-06T22:09:10.7922513Z Dict_test 2024-08-06T22:09:10.7923063Z Dimname_test 2024-08-06T22:09:10.7923595Z FileStoreTest 2024-08-06T22:09:10.7924124Z HashStoreTest 2024-08-06T22:09:10.7924490Z IListRef_test 2024-08-06T22:09:10.7924986Z KernelFunction_test 2024-08-06T22:09:10.7925566Z List_test 2024-08-06T22:09:10.7925918Z MaybeOwned_test 2024-08-06T22:09:10.7926269Z NamedTensor_test 2024-08-06T22:09:10.7926894Z ProcessGroupGlooAsyncTest 2024-08-06T22:09:10.7927632Z ProcessGroupGlooTest 2024-08-06T22:09:10.7928241Z ProcessGroupMPITest 2024-08-06T22:09:10.7928888Z ProcessGroupNCCLErrorsTest 2024-08-06T22:09:10.7929669Z ProcessGroupNCCLTest 2024-08-06T22:09:10.7930116Z StorageUtils_test 2024-08-06T22:09:10.7930433Z TCPStoreTest 2024-08-06T22:09:10.7930740Z aot_model_compiler_test 2024-08-06T22:09:10.7931068Z apply_utils_test 2024-08-06T22:09:10.7931368Z atest 2024-08-06T22:09:10.7931649Z backend_fallback_test 2024-08-06T22:09:10.7931951Z basic 2024-08-06T22:09:10.7932227Z broadcast_test 2024-08-06T22:09:10.7932527Z c10_Bitset_test 2024-08-06T22:09:10.7932851Z c10_CompileTimeFunctionPointer_test 2024-08-06T22:09:10.7934276Z c10_ConstexprCrc_test 2024-08-06T22:09:10.7934745Z c10_DeadlockDetection_test 2024-08-06T22:09:10.7935092Z c10_DeviceGuard_test 2024-08-06T22:09:10.7935415Z c10_Device_test 2024-08-06T22:09:10.7935713Z c10_DispatchKeySet_test 2024-08-06T22:09:10.7936058Z c10_Half_test 2024-08-06T22:09:10.7936375Z c10_InlineDeviceGuard_test 2024-08-06T22:09:10.7936724Z c10_InlineStreamGuard_test 2024-08-06T22:09:10.7937078Z c10_LeftRight_test 2024-08-06T22:09:10.7937406Z c10_Metaprogramming_test 2024-08-06T22:09:10.7937733Z c10_Scalar_test 2024-08-06T22:09:10.7938049Z c10_SizesAndStrides_test 2024-08-06T22:09:10.7938403Z c10_StreamGuard_test 2024-08-06T22:09:10.7938705Z c10_SymInt_test 2024-08-06T22:09:10.7939011Z c10_Synchronized_test 2024-08-06T22:09:10.7939340Z c10_ThreadLocal_test 2024-08-06T22:09:10.7939646Z c10_TypeIndex_test 2024-08-06T22:09:10.7939954Z c10_TypeList_test 2024-08-06T22:09:10.7940314Z c10_TypeTraits_test 2024-08-06T22:09:10.7940620Z c10_accumulate_test 2024-08-06T22:09:10.7940945Z c10_bfloat16_test 2024-08-06T22:09:10.7941504Z c10_bit_cast_test 2024-08-06T22:09:10.7941804Z c10_complex_math_test 2024-08-06T22:09:10.7942126Z c10_complex_test 2024-08-06T22:09:10.7942425Z c10_cow_test 2024-08-06T22:09:10.7942734Z c10_cuda_CUDAAssertionsTest_1_var_test 2024-08-06T22:09:10.7943187Z c10_cuda_CUDAAssertionsTest_catches_stream 2024-08-06T22:09:10.7943737Z c10_cuda_CUDAAssertionsTest_catches_thread_and_block_and_device 2024-08-06T22:09:10.7944278Z c10_cuda_CUDAAssertionsTest_from_2_processes 2024-08-06T22:09:10.7944850Z c10_cuda_CUDAAssertionsTest_multiple_writes_from_blocks_and_threads 2024-08-06T22:09:10.7945524Z c10_cuda_CUDAAssertionsTest_multiple_writes_from_multiple_blocks 2024-08-06T22:09:10.7946125Z c10_cuda_CUDAAssertionsTest_multiple_writes_from_same_block 2024-08-06T22:09:10.7946613Z c10_cuda_CUDATest 2024-08-06T22:09:10.7946932Z c10_exception_test 2024-08-06T22:09:10.7947225Z c10_flags_test 2024-08-06T22:09:10.7947515Z c10_generic_math_test 2024-08-06T22:09:10.7947862Z c10_intrusive_ptr_benchmark 2024-08-06T22:09:10.7948214Z c10_intrusive_ptr_test 2024-08-06T22:09:10.7948547Z c10_irange_test 2024-08-06T22:09:10.7948848Z c10_lazy_test 2024-08-06T22:09:10.7949128Z c10_logging_test 2024-08-06T22:09:10.7949514Z c10_optional_test 2024-08-06T22:09:10.7949914Z c10_ordered_preserving_dict_test 2024-08-06T22:09:10.7950277Z c10_registry_test 2024-08-06T22:09:10.7950591Z c10_small_vector_test 2024-08-06T22:09:10.7950897Z c10_ssize_test 2024-08-06T22:09:10.7951210Z c10_string_util_test 2024-08-06T22:09:10.7951532Z c10_string_view_test 2024-08-06T22:09:10.7951843Z c10_tempfile_test 2024-08-06T22:09:10.7952154Z c10_typeid_test 2024-08-06T22:09:10.7952458Z cpu_allocator_test 2024-08-06T22:09:10.7952761Z cpu_generator_test 2024-08-06T22:09:10.7953096Z cpu_profiling_allocator_test 2024-08-06T22:09:10.7953453Z cpu_rng_test 2024-08-06T22:09:10.7953762Z cuda_allocatorTraceTracker_test 2024-08-06T22:09:10.7954157Z cuda_allocator_test 2024-08-06T22:09:10.7954484Z cuda_apply_test 2024-08-06T22:09:10.7954786Z cuda_atomic_ops_test 2024-08-06T22:09:10.7955135Z cuda_caching_host_allocator_test 2024-08-06T22:09:10.7955521Z cuda_complex_math_test 2024-08-06T22:09:10.7955838Z cuda_complex_test 2024-08-06T22:09:10.7956151Z cuda_cub_test 2024-08-06T22:09:10.7956430Z cuda_cudnn_test 2024-08-06T22:09:10.7956734Z cuda_device_test 2024-08-06T22:09:10.7957051Z cuda_distributions_test 2024-08-06T22:09:10.7957376Z cuda_dlconvertor_test 2024-08-06T22:09:10.7957711Z cuda_generator_test 2024-08-06T22:09:10.7958028Z cuda_half_test 2024-08-06T22:09:10.7958319Z cuda_integer_divider_test 2024-08-06T22:09:10.7958669Z cuda_optional_test 2024-08-06T22:09:10.7959004Z cuda_packedtensoraccessor_test 2024-08-06T22:09:10.7959377Z cuda_reportMemoryUsage_test 2024-08-06T22:09:10.7959740Z cuda_stream_test 2024-08-06T22:09:10.7960050Z cuda_vectorized_test 2024-08-06T22:09:10.7960370Z dispatch_key_set_test 2024-08-06T22:09:10.7960700Z dlconvertor_test 2024-08-06T22:09:10.7961010Z example_allreduce 2024-08-06T22:09:10.7961365Z extension_backend_test 2024-08-06T22:09:10.7961733Z half_test 2024-08-06T22:09:10.7962007Z inline_container_test 2024-08-06T22:09:10.7962332Z ivalue_test 2024-08-06T22:09:10.7962639Z kernel_function_legacy_test 2024-08-06T22:09:10.7962991Z kernel_function_test 2024-08-06T22:09:10.7963332Z kernel_lambda_legacy_test 2024-08-06T22:09:10.7963679Z kernel_lambda_test 2024-08-06T22:09:10.7964123Z kernel_stackbased_test 2024-08-06T22:09:10.7964759Z lazy_tensor_test 2024-08-06T22:09:10.7965115Z legacy_vmap_test 2024-08-06T22:09:10.7965404Z libc10.so 2024-08-06T22:09:10.7965694Z libc10_cuda.so 2024-08-06T22:09:10.7965996Z libc10d_cuda_test.so 2024-08-06T22:09:10.7966307Z libcaffe2_nvrtc.so 2024-08-06T22:09:10.7966615Z 'libmkldnn*' 2024-08-06T22:09:10.7966884Z 'libnccl*' 2024-08-06T22:09:10.7967166Z libtorch.so 2024-08-06T22:09:10.7967458Z libtorch_cpu.so 2024-08-06T22:09:10.7967747Z libtorch_cuda.so 2024-08-06T22:09:10.7968061Z libtorch_cuda_linalg.so 2024-08-06T22:09:10.7968418Z libtorch_global_deps.so 2024-08-06T22:09:10.7968747Z libtorch_python.so 2024-08-06T22:09:10.7969074Z libtorchbind_test.so 2024-08-06T22:09:10.7969424Z make_boxed_from_unboxed_functor_test 2024-08-06T22:09:10.7969806Z math_kernel_test 2024-08-06T22:09:10.7970123Z memory_format_test 2024-08-06T22:09:10.7970449Z memory_overlapping_test 2024-08-06T22:09:10.7970781Z mobile_memory_cleanup 2024-08-06T22:09:10.7971109Z native_test 2024-08-06T22:09:10.7971406Z op_allowlist_test 2024-08-06T22:09:10.7971709Z op_registration_test 2024-08-06T22:09:10.7972042Z operator_name_test 2024-08-06T22:09:10.7972336Z operators_test 2024-08-06T22:09:10.7972660Z packedtensoraccessor_test 2024-08-06T22:09:10.7973020Z parallel_benchmark 2024-08-06T22:09:10.7973308Z pow_test 2024-08-06T22:09:10.7973586Z protoc 2024-08-06T22:09:10.7973865Z protoc-3.13.0.0 2024-08-06T22:09:10.7974157Z quantized_test 2024-08-06T22:09:10.7974458Z reduce_ops_test 2024-08-06T22:09:10.7974774Z reportMemoryUsage_test 2024-08-06T22:09:10.7975097Z scalar_tensor_test 2024-08-06T22:09:10.7975408Z scalar_test 2024-08-06T22:09:10.7975708Z stride_properties_test 2024-08-06T22:09:10.7976028Z tensor_iterator_test 2024-08-06T22:09:10.7976345Z test_api 2024-08-06T22:09:10.7976681Z test_cpp_rpc 2024-08-06T22:09:10.7977032Z test_dist_autograd 2024-08-06T22:09:10.7977355Z test_edge_op_registration 2024-08-06T22:09:10.7977682Z test_jit 2024-08-06T22:09:10.7977954Z test_lazy 2024-08-06T22:09:10.7978233Z test_mobile_nnc 2024-08-06T22:09:10.7978517Z test_parallel 2024-08-06T22:09:10.7978814Z test_tensorexpr 2024-08-06T22:09:10.7979116Z thread_init_test 2024-08-06T22:09:10.7979403Z torch_shm_manager 2024-08-06T22:09:10.7979713Z tutorial_tensorexpr 2024-08-06T22:09:10.7980034Z type_ptr_test 2024-08-06T22:09:10.7980314Z type_test 2024-08-06T22:09:10.7980608Z undefined_tensor_test 2024-08-06T22:09:10.7980949Z vec_test_all_types_AVX2 2024-08-06T22:09:10.7981285Z vec_test_all_types_AVX512 2024-08-06T22:09:10.7981647Z vec_test_all_types_DEFAULT 2024-08-06T22:09:10.7981991Z verify_api_visibility 2024-08-06T22:09:10.7982322Z weakref_test 2024-08-06T22:09:10.7982616Z wrapdim_test 2024-08-06T22:09:10.7982901Z xla_tensor_test 2024-08-06T22:09:10.7983232Z + aten/tools/run_tests.sh build/bin 2024-08-06T22:09:10.7983638Z + set -e 2024-08-06T22:09:10.7983923Z ++ dirname aten/tools/run_tests.sh 2024-08-06T22:09:10.7984441Z + VALGRIND_SUP=/var/lib/jenkins/workspace/aten/tools/valgrind.sup 2024-08-06T22:09:10.7984979Z + export CPP_TESTS_DIR=build/bin 2024-08-06T22:09:10.7985346Z + CPP_TESTS_DIR=build/bin 2024-08-06T22:09:10.7985693Z + VALGRIND=ON 2024-08-06T22:09:10.7987802Z + python test/run_test.py --cpp --verbose -i cpp/basic cpp/atest cpp/scalar_test cpp/broadcast_test cpp/wrapdim_test cpp/apply_utils_test cpp/dlconvertor_test cpp/native_test cpp/scalar_tensor_test cpp/undefined_tensor_test cpp/extension_backend_test cpp/lazy_tensor_test cpp/tensor_iterator_test cpp/Dimname_test cpp/Dict_test cpp/NamedTensor_test cpp/cpu_generator_test cpp/legacy_vmap_test cpp/operators_test 2024-08-06T22:09:10.9290614Z /var/lib/jenkins/workspace/test/run_test.py:21: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html 2024-08-06T22:09:10.9291686Z import pkg_resources 2024-08-06T22:09:15.2793925Z Downloading https://ossci-metrics.s3.amazonaws.com/slow-tests.json to /var/lib/jenkins/workspace/test/.pytorch-slow-tests.json 2024-08-06T22:09:15.2795375Z Downloading https://ossci-metrics.s3.amazonaws.com/disabled-tests-condensed.json to /var/lib/jenkins/workspace/test/.pytorch-disabled-tests.json 2024-08-06T22:09:15.2928606Z Found test times from artifacts 2024-08-06T22:09:15.3457432Z Found test times from artifacts 2024-08-06T22:09:15.3472667Z Running 25% of tests based on TD 2024-08-06T22:09:15.3479026Z Running parallel tests on 2 processes 2024-08-06T22:09:15.3479833Z Name: tests to run (est. time: 0.0min) 2024-08-06T22:09:15.3480447Z Serial tests (0): 2024-08-06T22:09:15.3480787Z Parallel tests (5): 2024-08-06T22:09:15.3481262Z cpp/Dict_test 1/1 2024-08-06T22:09:15.3481653Z cpp/Dimname_test 1/1 2024-08-06T22:09:15.3482027Z cpp/NamedTensor_test 1/1 2024-08-06T22:09:15.3482556Z cpp/apply_utils_test 1/1 2024-08-06T22:09:15.3482967Z cpp/atest 1/1 2024-08-06T22:09:15.3483323Z Name: excluded (est. time: 0.0min) 2024-08-06T22:09:15.3483715Z Serial tests (0): 2024-08-06T22:09:15.3484019Z Parallel tests (14): 2024-08-06T22:09:15.3484347Z cpp/basic 1/1 2024-08-06T22:09:15.3484668Z cpp/broadcast_test 1/1 2024-08-06T22:09:15.3485013Z cpp/cpu_generator_test 1/1 2024-08-06T22:09:15.3485407Z cpp/dlconvertor_test 1/1 2024-08-06T22:09:15.3485783Z cpp/extension_backend_test 1/1 2024-08-06T22:09:15.3486154Z cpp/lazy_tensor_test 1/1 2024-08-06T22:09:15.3486517Z cpp/legacy_vmap_test 1/1 2024-08-06T22:09:15.3486877Z cpp/native_test 1/1 2024-08-06T22:09:15.3487200Z cpp/operators_test 1/1 2024-08-06T22:09:15.3487559Z cpp/scalar_tensor_test 1/1 2024-08-06T22:09:15.3487909Z cpp/scalar_test 1/1 2024-08-06T22:09:15.3488262Z cpp/tensor_iterator_test 1/1 2024-08-06T22:09:15.3488651Z cpp/undefined_tensor_test 1/1 2024-08-06T22:09:15.3489018Z cpp/wrapdim_test 1/1 2024-08-06T22:09:15.3489751Z Starting test batch 'tests to run' 0.0 seconds after initiating testing 2024-08-06T22:09:15.3549065Z Running cpp/Dict_test 1/1 ... [2024-08-06 22:09:15.354550] 2024-08-06T22:09:15.3556735Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/Dict_test', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-13e8d588f452a2c6.xml', '-x', '--reruns=2'] ... [2024-08-06 22:09:15.355283] 2024-08-06T22:09:17.9761138Z 2024-08-06T22:09:17.9762905Z cpp/Dict_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.Dict_test_1.1_b8a93d126e35bc15_.log 2024-08-06T22:09:17.9764328Z 2024-08-06T22:09:17.9765573Z Running cpp/Dimname_test 1/1 ... [2024-08-06 22:09:17.976219] 2024-08-06T22:09:17.9773780Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/Dimname_test', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-650d23f330972634.xml', '-x', '--reruns=2'] ... [2024-08-06 22:09:17.976970] 2024-08-06T22:09:20.4471470Z 2024-08-06T22:09:20.4472861Z cpp/Dimname_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.Dimname_test_1.1_159a6d4beede0f99_.log 2024-08-06T22:09:20.4473637Z 2024-08-06T22:09:20.4473897Z Running cpp/NamedTensor_test 1/1 ... [2024-08-06 22:09:20.446923] 2024-08-06T22:09:20.4479023Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/NamedTensor_test', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-46ca08301864d329.xml', '-x', '--reruns=2'] ... [2024-08-06 22:09:20.447510] 2024-08-06T22:09:22.9672940Z 2024-08-06T22:09:22.9674525Z cpp/NamedTensor_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.NamedTensor_test_1.1_c2da3b1622e986d3_.log 2024-08-06T22:09:22.9675460Z 2024-08-06T22:09:22.9675745Z Running cpp/apply_utils_test 1/1 ... [2024-08-06 22:09:22.967016] 2024-08-06T22:09:22.9679166Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/apply_utils_test', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-de63f0ba675b5139.xml', '-x', '--reruns=2'] ... [2024-08-06 22:09:22.967499] 2024-08-06T22:09:25.4373009Z 2024-08-06T22:09:25.4374295Z cpp/apply_utils_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.apply_utils_test_1.1_6ce8fa5d18b43412_.log 2024-08-06T22:09:25.4375073Z 2024-08-06T22:09:25.4375288Z Running cpp/atest 1/1 ... [2024-08-06 22:09:25.437070] 2024-08-06T22:09:25.4380608Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/atest', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-f3b2e12e7c1f187a.xml', '-x', '--reruns=2'] ... [2024-08-06 22:09:25.437629] 2024-08-06T22:09:27.9080674Z 2024-08-06T22:09:27.9081832Z cpp/atest 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.atest_1.1_99622619a278fa84_.log 2024-08-06T22:09:27.9082537Z 2024-08-06T22:09:27.9089993Z Running cpp/Dict_test 1/1 ... [2024-08-06 22:09:27.908656] 2024-08-06T22:09:27.9092421Z Running cpp/Dimname_test 1/1 ... [2024-08-06 22:09:27.908878] 2024-08-06T22:09:27.9098988Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/Dict_test', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-5408deef3ec6127d.xml', '-x', '--reruns=2'] ... [2024-08-06 22:09:27.909450] 2024-08-06T22:09:27.9101513Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/Dimname_test', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-ae186a4a56d0f386.xml', '-x', '--reruns=2'] ... [2024-08-06 22:09:27.909568] 2024-08-06T22:09:31.7326685Z 2024-08-06T22:09:31.7329016Z cpp/Dimname_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.Dimname_test_1.1_6f9cae03abc80a0b_.log 2024-08-06T22:09:31.7330736Z 2024-08-06T22:09:35.5143526Z Running cpp/NamedTensor_test 1/1 ... [2024-08-06 22:09:35.513762] 2024-08-06T22:09:35.5148856Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/NamedTensor_test', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-a3c5c786a2735e95.xml', '-x', '--reruns=2'] ... [2024-08-06 22:09:35.514474] 2024-08-06T22:09:41.2432986Z 2024-08-06T22:09:41.2434981Z cpp/NamedTensor_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.NamedTensor_test_1.1_8d64a4fa755a2ae5_.log 2024-08-06T22:09:41.2436679Z 2024-08-06T22:09:45.0274610Z Running cpp/apply_utils_test 1/1 ... [2024-08-06 22:09:45.026985] 2024-08-06T22:09:45.0280043Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/apply_utils_test', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-8c90e77e148b2752.xml', '-x', '--reruns=2'] ... [2024-08-06 22:09:45.027629] 2024-08-06T22:09:45.7118696Z 2024-08-06T22:09:45.7120434Z cpp/Dict_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.Dict_test_1.1_b3dae9fe08b8f271_.log 2024-08-06T22:09:45.7121525Z 2024-08-06T22:09:49.4696495Z Running cpp/atest 1/1 ... [2024-08-06 22:09:49.469132] 2024-08-06T22:09:49.4702054Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/atest', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-0b59a7ae49ac25ca.xml', '-x', '--reruns=2'] ... [2024-08-06 22:09:49.469759] 2024-08-06T22:09:49.7023528Z 2024-08-06T22:09:49.7024565Z cpp/apply_utils_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.apply_utils_test_1.1_dc73b066b157f97a_.log 2024-08-06T22:09:49.7025409Z 2024-08-06T22:09:57.1504679Z 2024-08-06T22:09:57.1506140Z cpp/atest 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.atest_1.1_e3aa93d34d3c54a2_.log 2024-08-06T22:09:57.1506845Z 2024-08-06T22:09:58.1526855Z + run_if_exists tensor_interop_test 2024-08-06T22:09:58.1527355Z + local test_name=tensor_interop_test 2024-08-06T22:09:58.1527774Z + [[ -x build/bin/tensor_interop_test ]] 2024-08-06T22:09:58.1528251Z + echo 'Warning: tensor_interop_test does not exist.' 2024-08-06T22:09:58.1528755Z Warning: tensor_interop_test does not exist. 2024-08-06T22:09:58.1529184Z + run_if_exists cudnn_test 2024-08-06T22:09:58.1529544Z + local test_name=cudnn_test 2024-08-06T22:09:58.1529910Z + [[ -x build/bin/cudnn_test ]] 2024-08-06T22:09:58.1530310Z + echo 'Warning: cudnn_test does not exist.' 2024-08-06T22:09:58.1530754Z Warning: cudnn_test does not exist. 2024-08-06T22:09:58.1531155Z + run_if_exists cuda_generator_test 2024-08-06T22:09:58.1531550Z + local test_name=cuda_generator_test 2024-08-06T22:09:58.1531968Z + [[ -x build/bin/cuda_generator_test ]] 2024-08-06T22:09:58.1532509Z + python test/run_test.py --cpp --verbose -i cpp/cuda_generator_test 2024-08-06T22:09:58.2864349Z /var/lib/jenkins/workspace/test/run_test.py:21: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html 2024-08-06T22:09:58.2865395Z import pkg_resources 2024-08-06T22:10:02.6717903Z Downloading https://ossci-metrics.s3.amazonaws.com/slow-tests.json to /var/lib/jenkins/workspace/test/.pytorch-slow-tests.json 2024-08-06T22:10:02.6719347Z Downloading https://ossci-metrics.s3.amazonaws.com/disabled-tests-condensed.json to /var/lib/jenkins/workspace/test/.pytorch-disabled-tests.json 2024-08-06T22:10:02.6853387Z Found test times from artifacts 2024-08-06T22:10:02.7386022Z Found test times from artifacts 2024-08-06T22:10:02.7401060Z Running 25% of tests based on TD 2024-08-06T22:10:02.7404701Z Running parallel tests on 2 processes 2024-08-06T22:10:02.7405872Z Name: tests to run (est. time: 0.0min) 2024-08-06T22:10:02.7406415Z Serial tests (0): 2024-08-06T22:10:02.7406742Z Parallel tests (1): 2024-08-06T22:10:02.7407070Z cpp/cuda_generator_test 1/1 2024-08-06T22:10:02.7407451Z Name: excluded (est. time: 0.0min) 2024-08-06T22:10:02.7407827Z Serial tests (0): 2024-08-06T22:10:02.7408126Z Parallel tests (0): 2024-08-06T22:10:02.7408601Z Starting test batch 'tests to run' 0.0 seconds after initiating testing 2024-08-06T22:10:02.7468307Z Running cpp/cuda_generator_test 1/1 ... [2024-08-06 22:10:02.746446] 2024-08-06T22:10:02.7475047Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_generator_test', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-94c9344efbd6dc1b.xml', '-x', '--reruns=2'] ... [2024-08-06 22:10:02.747012] 2024-08-06T22:10:05.3677076Z 2024-08-06T22:10:05.3678169Z cpp/cuda_generator_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_generator_test_1.1_a341f51949e2695d_.log 2024-08-06T22:10:05.3679012Z 2024-08-06T22:10:05.9243904Z Running cpp/cuda_generator_test 1/1 ... [2024-08-06 22:10:05.923882] 2024-08-06T22:10:05.9249117Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_generator_test', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-a9cee4f64f224215.xml', '-x', '--reruns=2'] ... [2024-08-06 22:10:05.924504] 2024-08-06T22:10:12.7538325Z 2024-08-06T22:10:12.7540179Z cpp/cuda_generator_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_generator_test_1.1_63920664cba4ff58_.log 2024-08-06T22:10:12.7541005Z 2024-08-06T22:10:13.7047613Z + run_if_exists apply_test 2024-08-06T22:10:13.7048238Z + local test_name=apply_test 2024-08-06T22:10:13.7048630Z + [[ -x build/bin/apply_test ]] 2024-08-06T22:10:13.7049054Z + echo 'Warning: apply_test does not exist.' 2024-08-06T22:10:13.7049485Z Warning: apply_test does not exist. 2024-08-06T22:10:13.7049879Z + run_if_exists stream_test 2024-08-06T22:10:13.7050219Z + local test_name=stream_test 2024-08-06T22:10:13.7050589Z + [[ -x build/bin/stream_test ]] 2024-08-06T22:10:13.7050991Z + echo 'Warning: stream_test does not exist.' 2024-08-06T22:10:13.7051404Z Warning: stream_test does not exist. 2024-08-06T22:10:13.7051813Z + run_if_exists cuda_half_test 2024-08-06T22:10:13.7052186Z + local test_name=cuda_half_test 2024-08-06T22:10:13.7052549Z + [[ -x build/bin/cuda_half_test ]] 2024-08-06T22:10:13.7053059Z + python test/run_test.py --cpp --verbose -i cpp/cuda_half_test 2024-08-06T22:10:13.8364208Z /var/lib/jenkins/workspace/test/run_test.py:21: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html 2024-08-06T22:10:13.8365299Z import pkg_resources 2024-08-06T22:10:18.1672699Z Downloading https://ossci-metrics.s3.amazonaws.com/slow-tests.json to /var/lib/jenkins/workspace/test/.pytorch-slow-tests.json 2024-08-06T22:10:18.1674144Z Downloading https://ossci-metrics.s3.amazonaws.com/disabled-tests-condensed.json to /var/lib/jenkins/workspace/test/.pytorch-disabled-tests.json 2024-08-06T22:10:18.1809515Z Found test times from artifacts 2024-08-06T22:10:18.2341767Z Found test times from artifacts 2024-08-06T22:10:18.2358214Z Running 25% of tests based on TD 2024-08-06T22:10:18.2362135Z Running parallel tests on 2 processes 2024-08-06T22:10:18.2363073Z Name: tests to run (est. time: 0.0min) 2024-08-06T22:10:18.2363689Z Serial tests (0): 2024-08-06T22:10:18.2363998Z Parallel tests (1): 2024-08-06T22:10:18.2364336Z cpp/cuda_half_test 1/1 2024-08-06T22:10:18.2364706Z Name: excluded (est. time: 0.0min) 2024-08-06T22:10:18.2365081Z Serial tests (0): 2024-08-06T22:10:18.2365411Z Parallel tests (0): 2024-08-06T22:10:18.2365886Z Starting test batch 'tests to run' 0.0 seconds after initiating testing 2024-08-06T22:10:18.2425955Z Running cpp/cuda_half_test 1/1 ... [2024-08-06 22:10:18.242239] 2024-08-06T22:10:18.2433008Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_half_test', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-6dd991447a836f65.xml', '-x', '--reruns=2'] ... [2024-08-06 22:10:18.242929] 2024-08-06T22:10:20.8133571Z 2024-08-06T22:10:20.8134872Z cpp/cuda_half_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_half_test_1.1_8b27844259643beb_.log 2024-08-06T22:10:20.8135660Z 2024-08-06T22:10:21.3807233Z Running cpp/cuda_half_test 1/1 ... [2024-08-06 22:10:21.380134] 2024-08-06T22:10:21.3813198Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_half_test', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-732b5f356d37a9be.xml', '-x', '--reruns=2'] ... [2024-08-06 22:10:21.380877] 2024-08-06T22:10:24.6526977Z 2024-08-06T22:10:24.6528104Z cpp/cuda_half_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_half_test_1.1_7a878d97c4d54838_.log 2024-08-06T22:10:24.6528904Z 2024-08-06T22:10:25.6440242Z + run_if_exists cuda_vectorized_test 2024-08-06T22:10:25.6440774Z + local test_name=cuda_vectorized_test 2024-08-06T22:10:25.6441279Z + [[ -x build/bin/cuda_vectorized_test ]] 2024-08-06T22:10:25.6442013Z + python test/run_test.py --cpp --verbose -i cpp/cuda_vectorized_test 2024-08-06T22:10:25.7776244Z /var/lib/jenkins/workspace/test/run_test.py:21: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html 2024-08-06T22:10:25.7777276Z import pkg_resources 2024-08-06T22:10:30.1508525Z Downloading https://ossci-metrics.s3.amazonaws.com/slow-tests.json to /var/lib/jenkins/workspace/test/.pytorch-slow-tests.json 2024-08-06T22:10:30.1510118Z Downloading https://ossci-metrics.s3.amazonaws.com/disabled-tests-condensed.json to /var/lib/jenkins/workspace/test/.pytorch-disabled-tests.json 2024-08-06T22:10:30.1644109Z Found test times from artifacts 2024-08-06T22:10:30.2170512Z Found test times from artifacts 2024-08-06T22:10:30.2185487Z Running 25% of tests based on TD 2024-08-06T22:10:30.2189035Z Running parallel tests on 2 processes 2024-08-06T22:10:30.2189488Z Name: tests to run (est. time: 0.0min) 2024-08-06T22:10:30.2189873Z Serial tests (0): 2024-08-06T22:10:30.2190203Z Parallel tests (1): 2024-08-06T22:10:30.2190553Z cpp/cuda_vectorized_test 1/1 2024-08-06T22:10:30.2190929Z Name: excluded (est. time: 0.0min) 2024-08-06T22:10:30.2191305Z Serial tests (0): 2024-08-06T22:10:30.2191614Z Parallel tests (0): 2024-08-06T22:10:30.2192102Z Starting test batch 'tests to run' 0.0 seconds after initiating testing 2024-08-06T22:10:30.2253010Z Running cpp/cuda_vectorized_test 1/1 ... [2024-08-06 22:10:30.224897] 2024-08-06T22:10:30.2258777Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_vectorized_test', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-8f34bee56aa72227.xml', '-x', '--reruns=2'] ... [2024-08-06 22:10:30.225458] 2024-08-06T22:10:32.8462587Z 2024-08-06T22:10:32.8463915Z cpp/cuda_vectorized_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_vectorized_test_1.1_899f90a4ad4cc144_.log 2024-08-06T22:10:32.8464745Z 2024-08-06T22:10:33.3739998Z Running cpp/cuda_vectorized_test 1/1 ... [2024-08-06 22:10:33.373470] 2024-08-06T22:10:33.3746506Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_vectorized_test', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-63dc647abde41ca8.xml', '-x', '--reruns=2'] ... [2024-08-06 22:10:33.374078] 2024-08-06T22:10:36.7461422Z 2024-08-06T22:10:36.7463343Z cpp/cuda_vectorized_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_vectorized_test_1.1_1c80927537e8ad2d_.log 2024-08-06T22:10:36.7465074Z 2024-08-06T22:10:37.7289943Z + run_if_exists cuda_distributions_test 2024-08-06T22:10:37.7290474Z + local test_name=cuda_distributions_test 2024-08-06T22:10:37.7290943Z + [[ -x build/bin/cuda_distributions_test ]] 2024-08-06T22:10:37.7291508Z + python test/run_test.py --cpp --verbose -i cpp/cuda_distributions_test 2024-08-06T22:10:37.8604529Z /var/lib/jenkins/workspace/test/run_test.py:21: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html 2024-08-06T22:10:37.8605632Z import pkg_resources 2024-08-06T22:10:42.2160967Z Downloading https://ossci-metrics.s3.amazonaws.com/slow-tests.json to /var/lib/jenkins/workspace/test/.pytorch-slow-tests.json 2024-08-06T22:10:42.2162695Z Downloading https://ossci-metrics.s3.amazonaws.com/disabled-tests-condensed.json to /var/lib/jenkins/workspace/test/.pytorch-disabled-tests.json 2024-08-06T22:10:42.2296076Z Found test times from artifacts 2024-08-06T22:10:42.2822036Z Found test times from artifacts 2024-08-06T22:10:42.2837328Z Running 25% of tests based on TD 2024-08-06T22:10:42.2840749Z Running parallel tests on 2 processes 2024-08-06T22:10:42.2841180Z Name: tests to run (est. time: 0.0min) 2024-08-06T22:10:42.2841579Z Serial tests (0): 2024-08-06T22:10:42.2841908Z Parallel tests (1): 2024-08-06T22:10:42.2842247Z cpp/cuda_distributions_test 1/1 2024-08-06T22:10:42.2842659Z Name: excluded (est. time: 0.0min) 2024-08-06T22:10:42.2843041Z Serial tests (0): 2024-08-06T22:10:42.2843348Z Parallel tests (0): 2024-08-06T22:10:42.2843826Z Starting test batch 'tests to run' 0.0 seconds after initiating testing 2024-08-06T22:10:42.2904291Z Running cpp/cuda_distributions_test 1/1 ... [2024-08-06 22:10:42.289991] 2024-08-06T22:10:42.2910056Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_distributions_test', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-4f7ca3755400e7a7.xml', '-x', '--reruns=2'] ... [2024-08-06 22:10:42.290593] 2024-08-06T22:10:44.8610562Z 2024-08-06T22:10:44.8611790Z cpp/cuda_distributions_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_distributions_test_1.1_2a7266db1843eeaf_.log 2024-08-06T22:10:44.8612663Z 2024-08-06T22:10:45.4960382Z Running cpp/cuda_distributions_test 1/1 ... [2024-08-06 22:10:45.495427] 2024-08-06T22:10:45.4967032Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_distributions_test', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-9d2a11d8a6228a31.xml', '-x', '--reruns=2'] ... [2024-08-06 22:10:45.496103] 2024-08-06T22:10:49.7698311Z 2024-08-06T22:10:49.7699856Z cpp/cuda_distributions_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_distributions_test_1.1_5a7d63b3b0a942b3_.log 2024-08-06T22:10:49.7701010Z 2024-08-06T22:10:50.7651309Z + run_if_exists cuda_optional_test 2024-08-06T22:10:50.7651944Z + local test_name=cuda_optional_test 2024-08-06T22:10:50.7652362Z + [[ -x build/bin/cuda_optional_test ]] 2024-08-06T22:10:50.7653173Z + python test/run_test.py --cpp --verbose -i cpp/cuda_optional_test 2024-08-06T22:10:50.8977835Z /var/lib/jenkins/workspace/test/run_test.py:21: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html 2024-08-06T22:10:50.8978884Z import pkg_resources 2024-08-06T22:10:55.2610749Z Downloading https://ossci-metrics.s3.amazonaws.com/slow-tests.json to /var/lib/jenkins/workspace/test/.pytorch-slow-tests.json 2024-08-06T22:10:55.2612210Z Downloading https://ossci-metrics.s3.amazonaws.com/disabled-tests-condensed.json to /var/lib/jenkins/workspace/test/.pytorch-disabled-tests.json 2024-08-06T22:10:55.2745974Z Found test times from artifacts 2024-08-06T22:10:55.3272817Z Found test times from artifacts 2024-08-06T22:10:55.3288377Z Running 25% of tests based on TD 2024-08-06T22:10:55.3291854Z Running parallel tests on 2 processes 2024-08-06T22:10:55.3292309Z Name: tests to run (est. time: 0.0min) 2024-08-06T22:10:55.3292690Z Serial tests (0): 2024-08-06T22:10:55.3293015Z Parallel tests (1): 2024-08-06T22:10:55.3293360Z cpp/cuda_optional_test 1/1 2024-08-06T22:10:55.3293726Z Name: excluded (est. time: 0.0min) 2024-08-06T22:10:55.3294105Z Serial tests (0): 2024-08-06T22:10:55.3294421Z Parallel tests (0): 2024-08-06T22:10:55.3294877Z Starting test batch 'tests to run' 0.0 seconds after initiating testing 2024-08-06T22:10:55.3354771Z Running cpp/cuda_optional_test 1/1 ... [2024-08-06 22:10:55.335108] 2024-08-06T22:10:55.3360682Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_optional_test', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-866e7c4cc74bc259.xml', '-x', '--reruns=2'] ... [2024-08-06 22:10:55.335666] 2024-08-06T22:10:57.9565696Z 2024-08-06T22:10:57.9566762Z cpp/cuda_optional_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_optional_test_1.1_13f81acd68833150_.log 2024-08-06T22:10:57.9567567Z 2024-08-06T22:10:58.4907244Z Running cpp/cuda_optional_test 1/1 ... [2024-08-06 22:10:58.490170] 2024-08-06T22:10:58.4913545Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_optional_test', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-749240509acd8fc1.xml', '-x', '--reruns=2'] ... [2024-08-06 22:10:58.490919] 2024-08-06T22:11:01.6127817Z 2024-08-06T22:11:01.6129486Z cpp/cuda_optional_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_optional_test_1.1_1f16a7523e5d8b3d_.log 2024-08-06T22:11:01.6130642Z 2024-08-06T22:11:02.6270619Z + run_if_exists cuda_tensor_interop_test 2024-08-06T22:11:02.6271157Z + local test_name=cuda_tensor_interop_test 2024-08-06T22:11:02.6271634Z + [[ -x build/bin/cuda_tensor_interop_test ]] 2024-08-06T22:11:02.6272142Z + echo 'Warning: cuda_tensor_interop_test does not exist.' 2024-08-06T22:11:02.6272659Z Warning: cuda_tensor_interop_test does not exist. 2024-08-06T22:11:02.6273115Z + run_if_exists cuda_complex_test 2024-08-06T22:11:02.6273509Z + local test_name=cuda_complex_test 2024-08-06T22:11:02.6273899Z + [[ -x build/bin/cuda_complex_test ]] 2024-08-06T22:11:02.6274431Z + python test/run_test.py --cpp --verbose -i cpp/cuda_complex_test 2024-08-06T22:11:02.7590288Z /var/lib/jenkins/workspace/test/run_test.py:21: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html 2024-08-06T22:11:02.7591320Z import pkg_resources 2024-08-06T22:11:07.1210389Z Downloading https://ossci-metrics.s3.amazonaws.com/slow-tests.json to /var/lib/jenkins/workspace/test/.pytorch-slow-tests.json 2024-08-06T22:11:07.1212267Z Downloading https://ossci-metrics.s3.amazonaws.com/disabled-tests-condensed.json to /var/lib/jenkins/workspace/test/.pytorch-disabled-tests.json 2024-08-06T22:11:07.1345822Z Found test times from artifacts 2024-08-06T22:11:07.1874287Z Found test times from artifacts 2024-08-06T22:11:07.1889102Z Running 25% of tests based on TD 2024-08-06T22:11:07.1892628Z Running parallel tests on 2 processes 2024-08-06T22:11:07.1893056Z Name: tests to run (est. time: 0.0min) 2024-08-06T22:11:07.1893459Z Serial tests (0): 2024-08-06T22:11:07.1893786Z Parallel tests (1): 2024-08-06T22:11:07.1894117Z cpp/cuda_complex_test 1/1 2024-08-06T22:11:07.1894502Z Name: excluded (est. time: 0.0min) 2024-08-06T22:11:07.1894883Z Serial tests (0): 2024-08-06T22:11:07.1895183Z Parallel tests (0): 2024-08-06T22:11:07.1895673Z Starting test batch 'tests to run' 0.0 seconds after initiating testing 2024-08-06T22:11:07.1957013Z Running cpp/cuda_complex_test 1/1 ... [2024-08-06 22:11:07.195309] 2024-08-06T22:11:07.1962734Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_complex_test', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-578e36301e10b5c3.xml', '-x', '--reruns=2'] ... [2024-08-06 22:11:07.195879] 2024-08-06T22:11:09.8164824Z 2024-08-06T22:11:09.8166113Z cpp/cuda_complex_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_complex_test_1.1_d8c71e4d1b8456a5_.log 2024-08-06T22:11:09.8166934Z 2024-08-06T22:11:10.3920827Z Running cpp/cuda_complex_test 1/1 ... [2024-08-06 22:11:10.391553] 2024-08-06T22:11:10.3927103Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_complex_test', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-be9f41af8783d60b.xml', '-x', '--reruns=2'] ... [2024-08-06 22:11:10.392247] 2024-08-06T22:11:16.9202066Z 2024-08-06T22:11:16.9203368Z cpp/cuda_complex_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_complex_test_1.1_0a521ef059b08c5a_.log 2024-08-06T22:11:16.9204725Z 2024-08-06T22:11:17.8502230Z + run_if_exists cuda_complex_math_test 2024-08-06T22:11:17.8502736Z + local test_name=cuda_complex_math_test 2024-08-06T22:11:17.8503189Z + [[ -x build/bin/cuda_complex_math_test ]] 2024-08-06T22:11:17.8503761Z + python test/run_test.py --cpp --verbose -i cpp/cuda_complex_math_test 2024-08-06T22:11:17.9819222Z /var/lib/jenkins/workspace/test/run_test.py:21: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html 2024-08-06T22:11:17.9821313Z import pkg_resources 2024-08-06T22:11:22.3721458Z Downloading https://ossci-metrics.s3.amazonaws.com/slow-tests.json to /var/lib/jenkins/workspace/test/.pytorch-slow-tests.json 2024-08-06T22:11:22.3723004Z Downloading https://ossci-metrics.s3.amazonaws.com/disabled-tests-condensed.json to /var/lib/jenkins/workspace/test/.pytorch-disabled-tests.json 2024-08-06T22:11:22.3858283Z Found test times from artifacts 2024-08-06T22:11:22.4390418Z Found test times from artifacts 2024-08-06T22:11:22.4415887Z Running 25% of tests based on TD 2024-08-06T22:11:22.4416405Z Running parallel tests on 2 processes 2024-08-06T22:11:22.4416845Z Name: tests to run (est. time: 0.0min) 2024-08-06T22:11:22.4417248Z Serial tests (0): 2024-08-06T22:11:22.4417559Z Parallel tests (1): 2024-08-06T22:11:22.4417911Z cpp/cuda_complex_math_test 1/1 2024-08-06T22:11:22.4418293Z Name: excluded (est. time: 0.0min) 2024-08-06T22:11:22.4418672Z Serial tests (0): 2024-08-06T22:11:22.4418989Z Parallel tests (0): 2024-08-06T22:11:22.4419448Z Starting test batch 'tests to run' 0.0 seconds after initiating testing 2024-08-06T22:11:22.4476153Z Running cpp/cuda_complex_math_test 1/1 ... [2024-08-06 22:11:22.447199] 2024-08-06T22:11:22.4482878Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_complex_math_test', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-b8a86eb29b4318af.xml', '-x', '--reruns=2'] ... [2024-08-06 22:11:22.447854] 2024-08-06T22:11:25.1189632Z 2024-08-06T22:11:25.1191003Z cpp/cuda_complex_math_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_complex_math_test_1.1_9833f3c0d8743532_.log 2024-08-06T22:11:25.1191863Z 2024-08-06T22:11:25.6465091Z Running cpp/cuda_complex_math_test 1/1 ... [2024-08-06 22:11:25.645976] 2024-08-06T22:11:25.6470386Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_complex_math_test', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-e59d3f9e832e6b94.xml', '-x', '--reruns=2'] ... [2024-08-06 22:11:25.646585] 2024-08-06T22:11:40.5929417Z 2024-08-06T22:11:40.5930611Z cpp/cuda_complex_math_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_complex_math_test_1.1_9f6b435f13a3abac_.log 2024-08-06T22:11:40.5931468Z 2024-08-06T22:11:41.6104905Z + run_if_exists cuda_cub_test 2024-08-06T22:11:41.6105496Z + local test_name=cuda_cub_test 2024-08-06T22:11:41.6105877Z + [[ -x build/bin/cuda_cub_test ]] 2024-08-06T22:11:41.6106386Z + python test/run_test.py --cpp --verbose -i cpp/cuda_cub_test 2024-08-06T22:11:41.7415847Z /var/lib/jenkins/workspace/test/run_test.py:21: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html 2024-08-06T22:11:41.7416893Z import pkg_resources 2024-08-06T22:11:46.1132885Z Downloading https://ossci-metrics.s3.amazonaws.com/slow-tests.json to /var/lib/jenkins/workspace/test/.pytorch-slow-tests.json 2024-08-06T22:11:46.1134350Z Downloading https://ossci-metrics.s3.amazonaws.com/disabled-tests-condensed.json to /var/lib/jenkins/workspace/test/.pytorch-disabled-tests.json 2024-08-06T22:11:46.1268984Z Found test times from artifacts 2024-08-06T22:11:46.1806807Z Found test times from artifacts 2024-08-06T22:11:46.1823171Z Running 25% of tests based on TD 2024-08-06T22:11:46.1827221Z Running parallel tests on 2 processes 2024-08-06T22:11:46.1827663Z Name: tests to run (est. time: 0.0min) 2024-08-06T22:11:46.1828049Z Serial tests (0): 2024-08-06T22:11:46.1828379Z Parallel tests (1): 2024-08-06T22:11:46.1828715Z cpp/cuda_cub_test 1/1 2024-08-06T22:11:46.1829057Z Name: excluded (est. time: 0.0min) 2024-08-06T22:11:46.1830142Z Serial tests (0): 2024-08-06T22:11:46.1830467Z Parallel tests (0): 2024-08-06T22:11:46.1830928Z Starting test batch 'tests to run' 0.0 seconds after initiating testing 2024-08-06T22:11:46.1890555Z Running cpp/cuda_cub_test 1/1 ... [2024-08-06 22:11:46.188661] 2024-08-06T22:11:46.1896533Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_cub_test', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-566ac99c71cc7149.xml', '-x', '--reruns=2'] ... [2024-08-06 22:11:46.189233] 2024-08-06T22:11:48.8095814Z 2024-08-06T22:11:48.8096981Z cpp/cuda_cub_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_cub_test_1.1_e3f50c3bd7708007_.log 2024-08-06T22:11:48.8097748Z 2024-08-06T22:11:49.4072253Z Running cpp/cuda_cub_test 1/1 ... [2024-08-06 22:11:49.406672] 2024-08-06T22:11:49.4078749Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_cub_test', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-4731d6b6d01c750e.xml', '-x', '--reruns=2'] ... [2024-08-06 22:11:49.407435] 2024-08-06T22:11:53.2308741Z 2024-08-06T22:11:53.2309718Z cpp/cuda_cub_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_cub_test_1.1_aefe592eb78e04d1_.log 2024-08-06T22:11:53.2310506Z 2024-08-06T22:11:54.1903009Z + run_if_exists cuda_atomic_ops_test 2024-08-06T22:11:54.1903495Z + local test_name=cuda_atomic_ops_test 2024-08-06T22:11:54.1904339Z + [[ -x build/bin/cuda_atomic_ops_test ]] 2024-08-06T22:11:54.1904967Z + python test/run_test.py --cpp --verbose -i cpp/cuda_atomic_ops_test 2024-08-06T22:11:54.3215567Z /var/lib/jenkins/workspace/test/run_test.py:21: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html 2024-08-06T22:11:54.3216626Z import pkg_resources 2024-08-06T22:11:58.6361583Z Downloading https://ossci-metrics.s3.amazonaws.com/slow-tests.json to /var/lib/jenkins/workspace/test/.pytorch-slow-tests.json 2024-08-06T22:11:58.6363310Z Downloading https://ossci-metrics.s3.amazonaws.com/disabled-tests-condensed.json to /var/lib/jenkins/workspace/test/.pytorch-disabled-tests.json 2024-08-06T22:11:58.6497064Z Found test times from artifacts 2024-08-06T22:11:58.7026648Z Found test times from artifacts 2024-08-06T22:11:58.7042245Z Running 25% of tests based on TD 2024-08-06T22:11:58.7045417Z Running parallel tests on 2 processes 2024-08-06T22:11:58.7046077Z Name: tests to run (est. time: 0.0min) 2024-08-06T22:11:58.7046699Z Serial tests (0): 2024-08-06T22:11:58.7047124Z Parallel tests (1): 2024-08-06T22:11:58.7047449Z cpp/cuda_atomic_ops_test 1/1 2024-08-06T22:11:58.7047836Z Name: excluded (est. time: 0.0min) 2024-08-06T22:11:58.7048213Z Serial tests (0): 2024-08-06T22:11:58.7048511Z Parallel tests (0): 2024-08-06T22:11:58.7048986Z Starting test batch 'tests to run' 0.0 seconds after initiating testing 2024-08-06T22:11:58.7110251Z Running cpp/cuda_atomic_ops_test 1/1 ... [2024-08-06 22:11:58.710629] 2024-08-06T22:11:58.7116847Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_atomic_ops_test', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-ff4928e6ad8174d5.xml', '-x', '--reruns=2'] ... [2024-08-06 22:11:58.711276] 2024-08-06T22:12:01.2865656Z 2024-08-06T22:12:01.2867173Z cpp/cuda_atomic_ops_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_atomic_ops_test_1.1_638a5c21685dd3c4_.log 2024-08-06T22:12:01.2868288Z 2024-08-06T22:12:01.9097320Z Running cpp/cuda_atomic_ops_test 1/1 ... [2024-08-06 22:12:01.909248] 2024-08-06T22:12:01.9104733Z Executing ['pytest', '/var/lib/jenkins/workspace/build/bin/cuda_atomic_ops_test', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-05705466277ac539.xml', '-x', '--reruns=2'] ... [2024-08-06 22:12:01.910024] 2024-08-06T22:12:06.1841917Z 2024-08-06T22:12:06.1843349Z cpp/cuda_atomic_ops_test 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.cuda_atomic_ops_test_1.1_1e8239c3859eb2b3_.log 2024-08-06T22:12:06.1844191Z 2024-08-06T22:12:07.1521800Z + '[' ON == ON ']' 2024-08-06T22:12:07.1523067Z + valgrind --suppressions=/var/lib/jenkins/workspace/aten/tools/valgrind.sup --error-exitcode=1 build/bin/basic '--gtest_filter=-*CUDA' 2024-08-06T22:12:07.1659932Z ==16059== Memcheck, a memory error detector 2024-08-06T22:12:07.1660527Z ==16059== Copyright (C) 2002-2022, and GNU GPL'd, by Julian Seward et al. 2024-08-06T22:12:07.1661242Z ==16059== Using Valgrind-3.20.0 and LibVEX; rerun with -h for copyright info 2024-08-06T22:12:07.1661866Z ==16059== Command: build/bin/basic --gtest_filter=-*CUDA 2024-08-06T22:12:07.1662311Z ==16059== 2024-08-06T22:12:14.6645793Z ==16059== Warning: set address range perms: large range [0x59cba000, 0x7bc84000) (defined) 2024-08-06T22:13:17.2146785Z Running main() from /var/lib/jenkins/workspace/third_party/googletest/googletest/src/gtest_main.cc 2024-08-06T22:13:17.2439772Z Note: Google Test filter = -*CUDA 2024-08-06T22:13:17.2517086Z [==========] Running 4 tests from 1 test suite. 2024-08-06T22:13:17.2538853Z [----------] Global test environment set-up. 2024-08-06T22:13:17.2576487Z [----------] 4 tests from BasicTest 2024-08-06T22:13:17.2601331Z [ RUN ] BasicTest.BasicTestCPU 2024-08-06T22:13:17.6352309Z ==16059== Warning: noted but unhandled ioctl 0x30000001 with no size/direction hints. 2024-08-06T22:13:17.6353219Z ==16059== This could cause spurious value errors to appear. 2024-08-06T22:13:17.6353902Z ==16059== See README_MISSING_SYSCALL_OR_IOCTL for guidance on writing a proper wrapper. 2024-08-06T22:13:17.6359323Z ==16059== Warning: noted but unhandled ioctl 0x4b with no size/direction hints. 2024-08-06T22:13:17.6360217Z ==16059== This could cause spurious value errors to appear. 2024-08-06T22:13:17.6360899Z ==16059== See README_MISSING_SYSCALL_OR_IOCTL for guidance on writing a proper wrapper. 2024-08-06T22:13:17.6372080Z ==16059== Warning: noted but unhandled ioctl 0x27 with no size/direction hints. 2024-08-06T22:13:17.6372864Z ==16059== This could cause spurious value errors to appear. 2024-08-06T22:13:17.6373530Z ==16059== See README_MISSING_SYSCALL_OR_IOCTL for guidance on writing a proper wrapper. 2024-08-06T22:13:17.8101292Z ==16059== Warning: noted but unhandled ioctl 0x25 with no size/direction hints. 2024-08-06T22:13:17.8102252Z ==16059== This could cause spurious value errors to appear. 2024-08-06T22:13:17.8103026Z ==16059== See README_MISSING_SYSCALL_OR_IOCTL for guidance on writing a proper wrapper. 2024-08-06T22:13:17.8224484Z ==16059== Warning: noted but unhandled ioctl 0x17 with no size/direction hints. 2024-08-06T22:13:17.8225147Z ==16059== This could cause spurious value errors to appear. 2024-08-06T22:13:17.8225824Z ==16059== See README_MISSING_SYSCALL_OR_IOCTL for guidance on writing a proper wrapper. 2024-08-06T22:13:17.8400727Z ==16059== Warning: set address range perms: large range [0x2000000000, 0x3f00000000) (noaccess) 2024-08-06T22:13:18.8617083Z ==16059== Warning: set address range perms: large range [0x200000000, 0x500000000) (noaccess) 2024-08-06T22:13:18.8703320Z ==16059== Warning: set address range perms: large range [0x85200000, 0xa51ff000) (noaccess) 2024-08-06T22:13:24.4224931Z 4376 ms 2024-08-06T22:13:25.4061270Z 82 ms 2024-08-06T22:13:26.4569023Z 109 ms 2024-08-06T22:13:39.5286021Z [ OK ] BasicTest.BasicTestCPU (22265 ms) 2024-08-06T22:13:39.5993781Z [ RUN ] BasicTest.BasicTestHalfCPU 2024-08-06T22:13:41.9854078Z 2260 ms 2024-08-06T22:13:42.0584819Z 67 ms 2024-08-06T22:13:42.1670314Z 105 ms 2024-08-06T22:13:42.2467990Z [ OK ] BasicTest.BasicTestHalfCPU (2576 ms) 2024-08-06T22:13:42.2470056Z [ RUN ] BasicTest.FactoryMethodsTest 2024-08-06T22:13:42.5222227Z ==16059== Warning: noted but unhandled ioctl 0x19 with no size/direction hints. 2024-08-06T22:13:42.5224645Z ==16059== This could cause spurious value errors to appear. 2024-08-06T22:13:42.5225382Z ==16059== See README_MISSING_SYSCALL_OR_IOCTL for guidance on writing a proper wrapper. 2024-08-06T22:13:43.2286362Z ==16059== Warning: noted but unhandled ioctl 0x49 with no size/direction hints. 2024-08-06T22:13:43.2287192Z ==16059== This could cause spurious value errors to appear. 2024-08-06T22:13:43.2287865Z ==16059== See README_MISSING_SYSCALL_OR_IOCTL for guidance on writing a proper wrapper. 2024-08-06T22:13:43.5929317Z ==16059== Warning: noted but unhandled ioctl 0x21 with no size/direction hints. 2024-08-06T22:13:43.5930012Z ==16059== This could cause spurious value errors to appear. 2024-08-06T22:13:43.5930944Z ==16059== See README_MISSING_SYSCALL_OR_IOCTL for guidance on writing a proper wrapper. 2024-08-06T22:13:45.1653424Z ==16059== Warning: noted but unhandled ioctl 0x1b with no size/direction hints. 2024-08-06T22:13:45.1654127Z ==16059== This could cause spurious value errors to appear. 2024-08-06T22:13:45.1654823Z ==16059== See README_MISSING_SYSCALL_OR_IOCTL for guidance on writing a proper wrapper. 2024-08-06T22:13:45.3782032Z ==16059== Warning: noted but unhandled ioctl 0x41 with no size/direction hints. 2024-08-06T22:13:45.3782809Z ==16059== This could cause spurious value errors to appear. 2024-08-06T22:13:45.3783509Z ==16059== See README_MISSING_SYSCALL_OR_IOCTL for guidance on writing a proper wrapper. 2024-08-06T22:13:46.5101963Z [ OK ] BasicTest.FactoryMethodsTest (4262 ms) 2024-08-06T22:13:46.5102621Z [ RUN ] BasicTest.BasicStdTestCPU 2024-08-06T22:13:46.6879088Z Simple example: called once 2024-08-06T22:13:46.6998610Z throw: call_once will retry 2024-08-06T22:13:46.7455347Z Didn't throw, call_once will not attempt again 2024-08-06T22:13:46.7488932Z [ OK ] BasicTest.BasicStdTestCPU (238 ms) 2024-08-06T22:13:46.7515539Z [----------] 4 tests from BasicTest (29489 ms total) 2024-08-06T22:13:46.7515915Z 2024-08-06T22:13:46.7528408Z [----------] Global test environment tear-down 2024-08-06T22:13:46.7564828Z [==========] 4 tests from 1 test suite ran. (29514 ms total) 2024-08-06T22:13:46.7581445Z [ PASSED ] 4 tests. 2024-08-06T22:13:50.5039673Z ==16059== 2024-08-06T22:13:50.5057717Z ==16059== HEAP SUMMARY: 2024-08-06T22:13:50.5058261Z ==16059== in use at exit: 13,097,921 bytes in 38,483 blocks 2024-08-06T22:13:50.5059180Z ==16059== total heap usage: 981,346 allocs, 942,863 frees, 262,466,099 bytes allocated 2024-08-06T22:13:50.5059837Z ==16059== 2024-08-06T22:13:51.7653890Z ==16059== LEAK SUMMARY: 2024-08-06T22:13:51.7654343Z ==16059== definitely lost: 288 bytes in 3 blocks 2024-08-06T22:13:51.7654839Z ==16059== indirectly lost: 192 bytes in 2 blocks 2024-08-06T22:13:51.7655333Z ==16059== possibly lost: 24,544 bytes in 164 blocks 2024-08-06T22:13:51.7655836Z ==16059== still reachable: 13,072,897 bytes in 38,314 blocks 2024-08-06T22:13:51.7656333Z ==16059== suppressed: 0 bytes in 0 blocks 2024-08-06T22:13:51.7656900Z ==16059== Rerun with --leak-check=full to see details of leaked memory 2024-08-06T22:13:51.7657387Z ==16059== 2024-08-06T22:13:51.7657827Z ==16059== For lists of detected and suppressed errors, rerun with: -s 2024-08-06T22:13:51.7658483Z ==16059== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 4 from 4) 2024-08-06T22:13:51.9248907Z + [[ -x build/bin/tensor_interop_test ]] 2024-08-06T22:13:51.9251550Z + [[ -n '' ]] 2024-08-06T22:13:51.9252003Z + assert_git_not_dirty 2024-08-06T22:13:51.9252392Z + [[ linux-focal-cuda12.1-py3.10-gcc9 != *rocm* ]] 2024-08-06T22:13:51.9252886Z + [[ linux-focal-cuda12.1-py3.10-gcc9 != *xla* ]] 2024-08-06T22:13:51.9259564Z ++ git status --porcelain 2024-08-06T22:13:51.9260595Z ++ grep -v '?? third_party' 2024-08-06T22:13:52.1497191Z ++ true 2024-08-06T22:13:52.1499106Z + git_status= 2024-08-06T22:13:52.1499464Z + [[ -n '' ]] 2024-08-06T22:13:52.1499929Z + test_libtorch 1 2024-08-06T22:13:52.1500230Z + local SHARD=1 2024-08-06T22:13:52.1500542Z + [[ default != \s\l\o\w ]] 2024-08-06T22:13:52.1501080Z + echo 'Testing libtorch' 2024-08-06T22:13:52.1501463Z Testing libtorch 2024-08-06T22:13:52.1502614Z + ln -sf /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libbackend_with_compiler.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/bin 2024-08-06T22:13:52.1515244Z + ln -sf /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libjitbackend_test.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/bin 2024-08-06T22:13:52.1526131Z + ln -sf /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libcaffe2_nvrtc.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/bin 2024-08-06T22:13:52.1539540Z + ln -sf /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libc10.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libc10_cuda.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libc10d_cuda_test.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/bin 2024-08-06T22:13:52.1551879Z + ln -sf /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libshm.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/bin 2024-08-06T22:13:52.1567253Z + ln -sf /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libtorch.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libtorch_cpu.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libtorch_cuda_linalg.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libtorch_global_deps.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libtorch_python.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libtorchbind_test.so /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/bin 2024-08-06T22:13:52.1579345Z + ln -sf '/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/lib/libnvfuser*' /opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/bin 2024-08-06T22:13:52.1591589Z + export CPP_TESTS_DIR=/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/bin 2024-08-06T22:13:52.1592367Z + CPP_TESTS_DIR=/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/bin 2024-08-06T22:13:52.1592935Z + [[ -z 1 ]] 2024-08-06T22:13:52.1593224Z + [[ 1 == \1 ]] 2024-08-06T22:13:52.1593515Z + test_libtorch_api 2024-08-06T22:13:52.1594096Z + MNIST_DIR=/var/lib/jenkins/workspace/test/cpp/api/mnist 2024-08-06T22:13:52.1594901Z + python tools/download_mnist.py --quiet -d /var/lib/jenkins/workspace/test/cpp/api/mnist 2024-08-06T22:13:52.2200819Z Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz ... 2024-08-06T22:13:52.3089682Z Failed to download (trying next): 2024-08-06T22:13:52.3090111Z HTTP Error 403: Forbidden 2024-08-06T22:13:52.3095162Z Downloading https://ossci-datasets.s3.amazonaws.com/mnist/train-images-idx3-ubyte.gz ... 2024-08-06T22:13:52.7297130Z Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz ... 2024-08-06T22:13:52.8095552Z Failed to download (trying next): 2024-08-06T22:13:52.8096128Z HTTP Error 403: Forbidden 2024-08-06T22:13:52.8101478Z Downloading https://ossci-datasets.s3.amazonaws.com/mnist/train-labels-idx1-ubyte.gz ... 2024-08-06T22:13:52.8484195Z Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz ... 2024-08-06T22:13:52.9293312Z Failed to download (trying next): 2024-08-06T22:13:52.9293871Z HTTP Error 403: Forbidden 2024-08-06T22:13:52.9299474Z Downloading https://ossci-datasets.s3.amazonaws.com/mnist/t10k-images-idx3-ubyte.gz ... 2024-08-06T22:13:53.0132924Z Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz ... 2024-08-06T22:13:53.0938675Z Failed to download (trying next): 2024-08-06T22:13:53.0939348Z HTTP Error 403: Forbidden 2024-08-06T22:13:53.0946461Z Downloading https://ossci-datasets.s3.amazonaws.com/mnist/t10k-labels-idx1-ubyte.gz ... 2024-08-06T22:13:53.1317802Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *asan* ]] 2024-08-06T22:13:53.1318610Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *slow-gradcheck* ]] 2024-08-06T22:13:53.1319086Z + OMP_NUM_THREADS=2 2024-08-06T22:13:53.1319573Z + TORCH_CPP_TEST_MNIST_PATH=/var/lib/jenkins/workspace/test/cpp/api/mnist 2024-08-06T22:13:53.1320491Z + python test/run_test.py --cpp --verbose -i cpp/test_api -k 'not IMethodTest' 2024-08-06T22:13:53.2618124Z /var/lib/jenkins/workspace/test/run_test.py:21: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html 2024-08-06T22:13:53.2619169Z import pkg_resources 2024-08-06T22:13:57.5226115Z Downloading https://ossci-metrics.s3.amazonaws.com/slow-tests.json to /var/lib/jenkins/workspace/test/.pytorch-slow-tests.json 2024-08-06T22:13:57.5227559Z Downloading https://ossci-metrics.s3.amazonaws.com/disabled-tests-condensed.json to /var/lib/jenkins/workspace/test/.pytorch-disabled-tests.json 2024-08-06T22:13:57.5362030Z Found test times from artifacts 2024-08-06T22:13:57.5888030Z Found test times from artifacts 2024-08-06T22:13:57.5903432Z Running 25% of tests based on TD 2024-08-06T22:13:57.5907508Z Running parallel tests on 2 processes 2024-08-06T22:13:57.5907937Z Name: tests to run (est. time: 0.0min) 2024-08-06T22:13:57.5908355Z Serial tests (0): 2024-08-06T22:13:57.5908692Z Parallel tests (1): 2024-08-06T22:13:57.5909006Z cpp/test_api 1/1 2024-08-06T22:13:57.5909347Z Name: excluded (est. time: 0.0min) 2024-08-06T22:13:57.5909729Z Serial tests (0): 2024-08-06T22:13:57.5910038Z Parallel tests (0): 2024-08-06T22:13:57.5910509Z Starting test batch 'tests to run' 0.0 seconds after initiating testing 2024-08-06T22:13:57.6009409Z Running cpp/test_api 1/1 ... [2024-08-06 22:13:57.600521] 2024-08-06T22:13:57.6015862Z Executing ['pytest', '/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/bin/test_api', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-cf09d745b78abb64.xml', '-k', 'not IMethodTest', '-x', '--reruns=2'] ... [2024-08-06 22:13:57.601143] 2024-08-06T22:14:00.5725003Z 2024-08-06T22:14:00.5726352Z cpp/test_api 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.test_api_1.1_061a79b2722d930b_.log 2024-08-06T22:14:00.5727111Z 2024-08-06T22:14:00.6700613Z Running cpp/test_api 1/1 ... [2024-08-06 22:14:00.669601] 2024-08-06T22:14:00.6707849Z Executing ['pytest', '/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/bin/test_api', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-1e7140df4a722398.xml', '-k', 'not IMethodTest', '-x', '--reruns=2'] ... [2024-08-06 22:14:00.670279] 2024-08-06T22:20:31.7962680Z 2024-08-06T22:20:31.7965808Z cpp/test_api 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.test_api_1.1_1533c031bc1e263b_.log 2024-08-06T22:20:31.7983282Z 2024-08-06T22:20:32.7422734Z + python test/run_test.py --cpp --verbose -i cpp/test_tensorexpr 2024-08-06T22:20:32.8737554Z /var/lib/jenkins/workspace/test/run_test.py:21: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html 2024-08-06T22:20:32.8739153Z import pkg_resources 2024-08-06T22:20:37.2086665Z Downloading https://ossci-metrics.s3.amazonaws.com/slow-tests.json to /var/lib/jenkins/workspace/test/.pytorch-slow-tests.json 2024-08-06T22:20:37.2088142Z Downloading https://ossci-metrics.s3.amazonaws.com/disabled-tests-condensed.json to /var/lib/jenkins/workspace/test/.pytorch-disabled-tests.json 2024-08-06T22:20:37.2223235Z Found test times from artifacts 2024-08-06T22:20:37.2748990Z Found test times from artifacts 2024-08-06T22:20:37.2764795Z Running 25% of tests based on TD 2024-08-06T22:20:37.2768572Z Running parallel tests on 2 processes 2024-08-06T22:20:37.2768997Z Name: tests to run (est. time: 0.0min) 2024-08-06T22:20:37.2769396Z Serial tests (0): 2024-08-06T22:20:37.2769722Z Parallel tests (1): 2024-08-06T22:20:37.2770053Z cpp/test_tensorexpr 1/1 2024-08-06T22:20:37.2770428Z Name: excluded (est. time: 0.0min) 2024-08-06T22:20:37.2770817Z Serial tests (0): 2024-08-06T22:20:37.2771124Z Parallel tests (0): 2024-08-06T22:20:37.2771890Z Starting test batch 'tests to run' 0.0 seconds after initiating testing 2024-08-06T22:20:37.2865814Z Running cpp/test_tensorexpr 1/1 ... [2024-08-06 22:20:37.286184] 2024-08-06T22:20:37.2872338Z Executing ['pytest', '/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/bin/test_tensorexpr', '-m', 'serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-653e4887ebf10e8c.xml', '-x', '--reruns=2'] ... [2024-08-06 22:20:37.286807] 2024-08-06T22:20:39.7567391Z 2024-08-06T22:20:39.7568388Z cpp/test_tensorexpr 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.test_tensorexpr_1.1_01cf64f8660069a7_.log 2024-08-06T22:20:39.7569174Z 2024-08-06T22:20:40.4130867Z Running cpp/test_tensorexpr 1/1 ... [2024-08-06 22:20:40.412608] 2024-08-06T22:20:40.4136303Z Executing ['pytest', '/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/bin/test_tensorexpr', '-m', 'not serial', '-v', '-vv', '-rfEX', '-n', '2', '--junit-xml-reruns', 'test-reports/python-pytest/test.run_test/test.run_test-595d30a92a33af94.xml', '-x', '--reruns=2'] ... [2024-08-06 22:20:40.413220] 2024-08-06T22:25:34.2054625Z 2024-08-06T22:25:34.2058363Z cpp/test_tensorexpr 1/1 was successful, full logs can be found in artifacts with path test/test-reports/cpp.test_tensorexpr_1.1_ea8e56300758860e_.log 2024-08-06T22:25:34.2070983Z 2024-08-06T22:25:35.2191505Z + [[ linux-focal-cuda12.1-py3.10-gcc9 != *android* ]] 2024-08-06T22:25:35.2192091Z + [[ linux-focal-cuda12.1-py3.10-gcc9 != *cuda* ]] 2024-08-06T22:25:35.2192530Z + [[ -z 1 ]] 2024-08-06T22:25:35.2192823Z + [[ 1 == \2 ]] 2024-08-06T22:25:35.2193128Z + assert_git_not_dirty 2024-08-06T22:25:35.2193502Z + [[ linux-focal-cuda12.1-py3.10-gcc9 != *rocm* ]] 2024-08-06T22:25:35.2193994Z + [[ linux-focal-cuda12.1-py3.10-gcc9 != *xla* ]] 2024-08-06T22:25:35.2197740Z ++ git status --porcelain 2024-08-06T22:25:35.2199449Z ++ grep -v '?? third_party' 2024-08-06T22:25:35.4474216Z ++ true 2024-08-06T22:25:35.4475591Z + git_status= 2024-08-06T22:25:35.4476212Z + [[ -n '' ]] 2024-08-06T22:25:35.4476615Z + [[ linux-focal-cuda12.1-py3.10-gcc9 == *xpu* ]] 2024-08-06T22:25:35.4477298Z + cleanup_workspace 2024-08-06T22:25:35.4478227Z + echo 'sudo may print the following warning message that can be ignored. The chown command will still run.' 2024-08-06T22:25:35.4479321Z sudo may print the following warning message that can be ignored. The chown command will still run. 2024-08-06T22:25:35.4480094Z + echo ' sudo: setrlimit(RLIMIT_STACK): Operation not permitted' 2024-08-06T22:25:35.4480671Z sudo: setrlimit(RLIMIT_STACK): Operation not permitted 2024-08-06T22:25:35.4481343Z + echo 'For more details refer to https://github.com/sudo-project/sudo/issues/42' 2024-08-06T22:25:35.4482084Z For more details refer to https://github.com/sudo-project/sudo/issues/42 2024-08-06T22:25:35.4482666Z + sudo chown -R 1000 /var/lib/jenkins/workspace 2024-08-06T22:25:36.0510638Z ##[group]Run cat test/**/*_toprint.log || true 2024-08-06T22:25:36.0511166Z cat test/**/*_toprint.log || true 2024-08-06T22:25:36.0521879Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T22:25:36.0522363Z env: 2024-08-06T22:25:36.0522636Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:36.0523069Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:36.0523766Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:36.0524360Z ##[endgroup] 2024-08-06T22:25:36.0615926Z cat: 'test/**/*_toprint.log': No such file or directory 2024-08-06T22:25:36.0671902Z ##[group]Run kill "$MONITOR_SCRIPT_PID" 2024-08-06T22:25:36.0672409Z kill "$MONITOR_SCRIPT_PID" 2024-08-06T22:25:36.0680941Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T22:25:36.0681421Z env: 2024-08-06T22:25:36.0681693Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:36.0682119Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:36.0682796Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:36.0683578Z MONITOR_SCRIPT_PID: 205326 2024-08-06T22:25:36.0683931Z ##[endgroup] 2024-08-06T22:25:36.0902282Z Prepare all required actions 2024-08-06T22:25:36.0902853Z Getting action download info 2024-08-06T22:25:36.3100077Z Download action repository 'actions/upload-artifact@v3' (SHA:a8a3f3ad30e3422c9c7b888a15615d19a852ae32) 2024-08-06T22:25:36.5433444Z ##[group]Run ./.github/actions/upload-test-artifacts 2024-08-06T22:25:36.5433909Z with: 2024-08-06T22:25:36.5434393Z file-suffix: test-default-1-5-amz2023.linux.4xlarge.nvidia.gpu_28428065757 2024-08-06T22:25:36.5434982Z s3-bucket: gha-artifacts 2024-08-06T22:25:36.5435326Z env: 2024-08-06T22:25:36.5435612Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:36.5436025Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:36.5436708Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:36.5437318Z ##[endgroup] 2024-08-06T22:25:36.5498960Z ##[group]Run # Remove any previous test jsons if they exist 2024-08-06T22:25:36.5499596Z # Remove any previous test jsons if they exist 2024-08-06T22:25:36.5500079Z rm -f test-jsons-*.zip 2024-08-06T22:25:36.5500566Z zip -r "test-jsons-${FILE_SUFFIX}.zip" test -i '*.json' 2024-08-06T22:25:36.5511146Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T22:25:36.5511634Z env: 2024-08-06T22:25:36.5511926Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:36.5512346Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:36.5513044Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:36.5513842Z FILE_SUFFIX: test-default-1-5-amz2023.linux.4xlarge.nvidia.gpu_28428065757 2024-08-06T22:25:36.5514416Z ##[endgroup] 2024-08-06T22:25:36.5780108Z adding: test/allowlist_for_publicAPI.json (deflated 79%) 2024-08-06T22:25:36.5815529Z adding: test/benchmark_utils/callgrind_artifacts.json (deflated 92%) 2024-08-06T22:25:36.5816293Z adding: test/minioptest_failures_dict.json (deflated 70%) 2024-08-06T22:25:36.5824012Z adding: test/profiler/profiler_utils_mock_events.json (deflated 87%) 2024-08-06T22:25:36.5828061Z adding: test/test-reports/td_exclusions-b0eb0482dfc102c57139.json (deflated 81%) 2024-08-06T22:25:36.5829728Z adding: test/test-reports/td_exclusions-18c2096214c81bbf27fe.json (deflated 18%) 2024-08-06T22:25:36.5831085Z adding: test/test-reports/td_exclusions-16c1b5827c926332dcee.json (deflated 73%) 2024-08-06T22:25:36.5832092Z adding: test/test-reports/td_exclusions-06fa2d2de6795a8450c2.json (deflated 16%) 2024-08-06T22:25:36.5833099Z adding: test/test-reports/td_exclusions-05d84b5fdd84e0793b50.json (deflated 14%) 2024-08-06T22:25:36.5833887Z adding: test/test-reports/td_exclusions-a90c4462e77ec2f641f6.json (deflated 15%) 2024-08-06T22:25:36.5834684Z adding: test/test-reports/td_exclusions-605b7b964fc701355a78.json (deflated 14%) 2024-08-06T22:25:36.5835455Z adding: test/test-reports/td_exclusions-d541c00c86e69220d463.json (deflated 13%) 2024-08-06T22:25:36.5836264Z adding: test/test-reports/td_exclusions-2156a3a657c94cd5a4e1.json (deflated 14%) 2024-08-06T22:25:36.5837052Z adding: test/test-reports/td_exclusions-aa226ae56597f0d09511.json (deflated 14%) 2024-08-06T22:25:36.5837825Z adding: test/test-reports/td_exclusions-701e7bd6e58429499dd3.json (deflated 13%) 2024-08-06T22:25:36.5838613Z adding: test/test-reports/td_exclusions-0eb43be561745c4d6850.json (deflated 15%) 2024-08-06T22:25:36.5839397Z adding: test/test-reports/td_exclusions-3a18891e057b17c2debc.json (deflated 14%) 2024-08-06T22:25:36.5840086Z adding: test/.pytorch-slow-tests.json (deflated 66%) 2024-08-06T22:25:36.5848538Z adding: test/.pytorch-disabled-tests.json (deflated 89%) 2024-08-06T22:25:36.5894903Z ##[group]Run # Remove any previous test reports if they exist 2024-08-06T22:25:36.5895497Z # Remove any previous test reports if they exist 2024-08-06T22:25:36.5895991Z rm -f test-reports-*.zip 2024-08-06T22:25:36.5896658Z zip -r "test-reports-${FILE_SUFFIX}.zip" test -i '*.xml' -i '*.csv' 2024-08-06T22:25:36.5903510Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T22:25:36.5904614Z env: 2024-08-06T22:25:36.5905135Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:36.5905773Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:36.5906627Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:36.5907433Z FILE_SUFFIX: test-default-1-5-amz2023.linux.4xlarge.nvidia.gpu_28428065757 2024-08-06T22:25:36.5907985Z ##[endgroup] 2024-08-06T22:25:36.6157305Z adding: test/test-reports/python-pytest/test_native_mha/test_native_mha-8f97b6dbb12707b4.xml (deflated 93%) 2024-08-06T22:25:36.6243051Z adding: test/test-reports/python-pytest/test_nn/test_nn-29ea430cff4674d5.xml (deflated 97%) 2024-08-06T22:25:36.6260226Z adding: test/test-reports/python-pytest/test_overrides/test_overrides-634e79f12529ee89.xml (deflated 95%) 2024-08-06T22:25:36.6268831Z adding: test/test-reports/python-pytest/test_cpp_api_parity/test_cpp_api_parity-17943bda628f37e7.xml (deflated 93%) 2024-08-06T22:25:36.6289339Z adding: test/test-reports/python-pytest/test_torch/test_torch-177927e37460b514.xml (deflated 92%) 2024-08-06T22:25:36.6290503Z adding: test/test-reports/python-pytest/test_namedtuple_return_api/test_namedtuple_return_api-4010d6d8a8473b63.xml (deflated 72%) 2024-08-06T22:25:36.6291641Z adding: test/test-reports/python-pytest/test_decomp/test_decomp-872c2be514219d60.xml (deflated 28%) 2024-08-06T22:25:36.6292616Z adding: test/test-reports/python-pytest/test_decomp/test_decomp-27c5dbd9310bb45c.xml (deflated 28%) 2024-08-06T22:25:36.6293608Z adding: test/test-reports/python-pytest/test_decomp/test_decomp-a5fc1e0970bfa54e.xml (deflated 28%) 2024-08-06T22:25:36.6294593Z adding: test/test-reports/python-pytest/test_decomp/test_decomp-54fe4f3c9997b261.xml (deflated 28%) 2024-08-06T22:25:36.6299813Z adding: test/test-reports/python-pytest/test_decomp/test_decomp-7b40ddb166521fca.xml (deflated 91%) 2024-08-06T22:25:36.6307955Z adding: test/test-reports/python-pytest/test_decomp/test_decomp-c27fe0d555463c84.xml (deflated 91%) 2024-08-06T22:25:36.6314680Z adding: test/test-reports/python-pytest/test_decomp/test_decomp-a2ad9593e18decc5.xml (deflated 91%) 2024-08-06T22:25:36.6321817Z adding: test/test-reports/python-pytest/test_decomp/test_decomp-f99458d9d8b0d312.xml (deflated 90%) 2024-08-06T22:25:36.6322974Z adding: test/test-reports/python-pytest/test_model_exports_to_core_aten/test_model_exports_to_core_aten-d25096b13d21e6b8.xml (deflated 28%) 2024-08-06T22:25:36.6324331Z adding: test/test-reports/python-pytest/test_model_exports_to_core_aten/test_model_exports_to_core_aten-637172318d6fea86.xml (deflated 58%) 2024-08-06T22:25:36.6325571Z adding: test/test-reports/python-pytest/test_transformers/test_transformers-23dafe8606154b9a.xml (deflated 28%) 2024-08-06T22:25:36.7300926Z adding: test/test-reports/python-pytest/test_transformers/test_transformers-cd06bb4e6af626fa.xml (deflated 99%) 2024-08-06T22:25:36.7302116Z adding: test/test-reports/python-pytest/test_transformers/test_transformers-447fb59da2867f77.xml (deflated 35%) 2024-08-06T22:25:36.7443302Z adding: test/test-reports/python-pytest/test_transformers/test_transformers-00a63c3f2f793df3.xml (deflated 98%) 2024-08-06T22:25:36.7444455Z adding: test/test-reports/python-pytest/functorch.test_ops/functorch.test_ops-e985e4d463f22058.xml (deflated 28%) 2024-08-06T22:25:36.7474411Z adding: test/test-reports/python-pytest/functorch.test_ops/functorch.test_ops-fe5d880d022b3d26.xml (deflated 93%) 2024-08-06T22:25:36.7475532Z adding: test/test-reports/python-pytest/test_proxy_tensor/test_proxy_tensor-d78972e1d7f7adc9.xml (deflated 28%) 2024-08-06T22:25:36.7477785Z adding: test/test-reports/python-pytest/test_proxy_tensor/test_proxy_tensor-ea600b78075595db.xml (deflated 91%) 2024-08-06T22:25:36.7478848Z adding: test/test-reports/python-pytest/test_foreach/test_foreach-2a7d3a0f80fd30cd.xml (deflated 28%) 2024-08-06T22:25:36.7530408Z adding: test/test-reports/python-pytest/test_foreach/test_foreach-fe7acb4d2f76e19e.xml (deflated 96%) 2024-08-06T22:25:36.7531503Z adding: test/test-reports/python-pytest/test_ops_gradients/test_ops_gradients-40c8aa748a31913c.xml (deflated 29%) 2024-08-06T22:25:36.7786573Z adding: test/test-reports/python-pytest/test_ops_gradients/test_ops_gradients-bb455e23f2899e58.xml (deflated 99%) 2024-08-06T22:25:36.7787640Z adding: test/test-reports/python-pytest/test_meta/test_meta-869f55c6802619d5.xml (deflated 28%) 2024-08-06T22:25:36.7949521Z adding: test/test-reports/python-pytest/test_meta/test_meta-4bf43e9730804140.xml (deflated 96%) 2024-08-06T22:25:36.7950471Z adding: test/test-reports/python-pytest/test_ops/test_ops-cc4f5b5b5a9d0077.xml (deflated 28%) 2024-08-06T22:25:36.7951402Z adding: test/test-reports/python-pytest/test_ops/test_ops-03bbfaa721f1008f.xml (deflated 28%) 2024-08-06T22:25:36.8046415Z adding: test/test-reports/python-pytest/test_ops/test_ops-9fa6df64a3bd39a4.xml (deflated 96%) 2024-08-06T22:25:36.8146937Z adding: test/test-reports/python-pytest/test_ops/test_ops-f91858ef5948256b.xml (deflated 97%) 2024-08-06T22:25:36.8147986Z adding: test/test-reports/python-pytest/test_module_tracker/test_module_tracker-ffb0a73abd462989.xml (deflated 28%) 2024-08-06T22:25:36.8149144Z adding: test/test-reports/python-pytest/test_module_tracker/test_module_tracker-685edc7cdf8b0ace.xml (deflated 50%) 2024-08-06T22:25:36.8150300Z adding: test/test-reports/python-pytest/test_comparison_utils/test_comparison_utils-29a25b4d75708cd5.xml (deflated 28%) 2024-08-06T22:25:36.8151489Z adding: test/test-reports/python-pytest/test_comparison_utils/test_comparison_utils-1a6ba41ff2791aed.xml (deflated 71%) 2024-08-06T22:25:36.8152667Z adding: test/test-reports/python-pytest/functorch.test_vmap/functorch.test_vmap-524b2c039f8cd547.xml (deflated 28%) 2024-08-06T22:25:36.8212426Z adding: test/test-reports/python-pytest/functorch.test_vmap/functorch.test_vmap-e05d8356dc924e7f.xml (deflated 94%) 2024-08-06T22:25:36.8213603Z adding: test/test-reports/python-pytest/export.test_tools/export.test_tools-9deeff2d5d2c4785.xml (deflated 28%) 2024-08-06T22:25:36.8214911Z adding: test/test-reports/python-pytest/export.test_tools/export.test_tools-6e7ec7ce2a80c721.xml (deflated 48%) 2024-08-06T22:25:36.8216296Z adding: test/test-reports/python-pytest/inductor.test_torchinductor_dynamic_shapes/inductor.test_torchinductor_dynamic_shapes-a09587fad9cd17a5.xml (deflated 58%) 2024-08-06T22:25:36.8221292Z adding: test/test-reports/python-pytest/inductor.test_torchinductor_dynamic_shapes/inductor.test_torchinductor_dynamic_shapes-329d6f06b623df53.xml (deflated 93%) 2024-08-06T22:25:36.8222746Z adding: test/test-reports/python-pytest/torch_np.test_binary_ufuncs/torch_np.test_binary_ufuncs-ace69e9b17832050.xml (deflated 28%) 2024-08-06T22:25:36.8224033Z adding: test/test-reports/python-pytest/torch_np.test_binary_ufuncs/torch_np.test_binary_ufuncs-3c19cd3bf4e16078.xml (deflated 91%) 2024-08-06T22:25:36.8225430Z adding: test/test-reports/python-pytest/dynamo.test_inline_inbuilt_nn_modules/dynamo.test_inline_inbuilt_nn_modules-442cbe56842c129d.xml (deflated 28%) 2024-08-06T22:25:36.8283323Z adding: test/test-reports/python-pytest/dynamo.test_inline_inbuilt_nn_modules/dynamo.test_inline_inbuilt_nn_modules-ed910124e333eea8.xml (deflated 92%) 2024-08-06T22:25:36.8284726Z adding: test/test-reports/python-pytest/inductor.test_inductor_utils/inductor.test_inductor_utils-97b30b7544df6d2f.xml (deflated 28%) 2024-08-06T22:25:36.8286068Z adding: test/test-reports/python-pytest/inductor.test_inductor_utils/inductor.test_inductor_utils-b6de974c188e7891.xml (deflated 57%) 2024-08-06T22:25:36.8287372Z adding: test/test-reports/python-pytest/torch_np.test_unary_ufuncs/torch_np.test_unary_ufuncs-c00ac13877fc2d14.xml (deflated 28%) 2024-08-06T22:25:36.8288653Z adding: test/test-reports/python-pytest/torch_np.test_unary_ufuncs/torch_np.test_unary_ufuncs-3f6a117be198c367.xml (deflated 91%) 2024-08-06T22:25:36.8290237Z adding: test/test-reports/python-pytest/inductor.test_extension_backend/inductor.test_extension_backend-7a3b8d4703e88d84.xml (deflated 28%) 2024-08-06T22:25:36.8291663Z adding: test/test-reports/python-pytest/inductor.test_extension_backend/inductor.test_extension_backend-2c05de40649b01c1.xml (deflated 52%) 2024-08-06T22:25:36.8293034Z adding: test/test-reports/python-pytest/export.test_export_nonstrict/export.test_export_nonstrict-ab85802882ef27d8.xml (deflated 28%) 2024-08-06T22:25:36.8296545Z adding: test/test-reports/python-pytest/export.test_export_nonstrict/export.test_export_nonstrict-41216224d07ba592.xml (deflated 89%) 2024-08-06T22:25:36.8297968Z adding: test/test-reports/python-pytest/inductor.test_triton_extension_backend/inductor.test_triton_extension_backend-dc2b79891e25ec81.xml (deflated 28%) 2024-08-06T22:25:36.8299500Z adding: test/test-reports/python-pytest/inductor.test_triton_extension_backend/inductor.test_triton_extension_backend-b67798e97e1b2afe.xml (deflated 42%) 2024-08-06T22:25:36.8300945Z adding: test/test-reports/python-pytest/inductor.test_triton_kernels/inductor.test_triton_kernels-3e7dd92337f931b3.xml (deflated 28%) 2024-08-06T22:25:36.8302931Z adding: test/test-reports/python-pytest/inductor.test_triton_kernels/inductor.test_triton_kernels-bcc35002ee704a64.xml (deflated 97%) 2024-08-06T22:25:36.8304437Z adding: test/test-reports/python-pytest/dynamo.test_base_output/dynamo.test_base_output-eddfe4d5191b0b6a.xml (deflated 28%) 2024-08-06T22:25:36.8305707Z adding: test/test-reports/python-pytest/dynamo.test_base_output/dynamo.test_base_output-7a5b091c66a55f97.xml (deflated 82%) 2024-08-06T22:25:36.8306894Z adding: test/test-reports/python-pytest/export.test_serdes/export.test_serdes-e56ac2dc66597e32.xml (deflated 28%) 2024-08-06T22:25:36.8315732Z adding: test/test-reports/python-pytest/export.test_serdes/export.test_serdes-145d35c75abe1098.xml (deflated 89%) 2024-08-06T22:25:36.8316867Z adding: test/test-reports/python-pytest/dynamo.test_interop/dynamo.test_interop-a452a6502540e5ec.xml (deflated 28%) 2024-08-06T22:25:36.8318037Z adding: test/test-reports/python-pytest/dynamo.test_interop/dynamo.test_interop-7afd4244412107ad.xml (deflated 70%) 2024-08-06T22:25:36.8319423Z adding: test/test-reports/python-pytest/inductor.test_compile_worker/inductor.test_compile_worker-43ef31a7ded0b80e.xml (deflated 28%) 2024-08-06T22:25:36.8320768Z adding: test/test-reports/python-pytest/inductor.test_compile_worker/inductor.test_compile_worker-23a93f5904da5fc5.xml (deflated 63%) 2024-08-06T22:25:36.8322087Z adding: test/test-reports/python-pytest/dynamo.test_fx_passes_pre_grad/dynamo.test_fx_passes_pre_grad-83e6abf840d99341.xml (deflated 28%) 2024-08-06T22:25:36.8323418Z adding: test/test-reports/python-pytest/dynamo.test_fx_passes_pre_grad/dynamo.test_fx_passes_pre_grad-357e00d9984eb300.xml (deflated 35%) 2024-08-06T22:25:36.8324736Z adding: test/test-reports/python-pytest/inductor.test_flex_decoding/inductor.test_flex_decoding-d69367c84f9c30a2.xml (deflated 28%) 2024-08-06T22:25:36.8326073Z adding: test/test-reports/python-pytest/inductor.test_flex_decoding/inductor.test_flex_decoding-fb493faf2afe5b4b.xml (deflated 97%) 2024-08-06T22:25:36.8327365Z adding: test/test-reports/python-pytest/dynamo.test_skip_non_tensor/dynamo.test_skip_non_tensor-789f7be7e6d3a27d.xml (deflated 28%) 2024-08-06T22:25:36.8328655Z adding: test/test-reports/python-pytest/dynamo.test_skip_non_tensor/dynamo.test_skip_non_tensor-371f4239eecbce67.xml (deflated 75%) 2024-08-06T22:25:36.8329842Z adding: test/test-reports/python-pytest/test_matmul_cuda/test_matmul_cuda-1f0a5b7b0a2073de.xml (deflated 28%) 2024-08-06T22:25:36.8330928Z adding: test/test-reports/python-pytest/test_matmul_cuda/test_matmul_cuda-9dcef9721e99a500.xml (deflated 92%) 2024-08-06T22:25:36.8331995Z adding: test/test-reports/python-pytest/dynamo.test_sdpa/dynamo.test_sdpa-d7cc1edf467a193d.xml (deflated 28%) 2024-08-06T22:25:36.8333089Z adding: test/test-reports/python-pytest/dynamo.test_sdpa/dynamo.test_sdpa-e5655976c9f8cbb3.xml (deflated 69%) 2024-08-06T22:25:36.8334474Z adding: test/test-reports/python-pytest/inductor.test_debug_trace/inductor.test_debug_trace-2ab5565bb8bb178e.xml (deflated 28%) 2024-08-06T22:25:36.8335754Z adding: test/test-reports/python-pytest/inductor.test_debug_trace/inductor.test_debug_trace-5f55e0d5260bddad.xml (deflated 51%) 2024-08-06T22:25:36.8336891Z adding: test/test-reports/python-pytest/test_modules/test_modules-3f434163cd29a6d6.xml (deflated 28%) 2024-08-06T22:25:36.8473600Z adding: test/test-reports/python-pytest/test_modules/test_modules-a755f83fa86fc6f8.xml (deflated 99%) 2024-08-06T22:25:36.8474629Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-13e8d588f452a2c6.xml (deflated 29%) 2024-08-06T22:25:36.8475626Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-650d23f330972634.xml (deflated 29%) 2024-08-06T22:25:36.8476635Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-46ca08301864d329.xml (deflated 29%) 2024-08-06T22:25:36.8477683Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-de63f0ba675b5139.xml (deflated 29%) 2024-08-06T22:25:36.8478683Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-f3b2e12e7c1f187a.xml (deflated 29%) 2024-08-06T22:25:36.8479711Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-ae186a4a56d0f386.xml (deflated 58%) 2024-08-06T22:25:36.8480725Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-a3c5c786a2735e95.xml (deflated 73%) 2024-08-06T22:25:36.8481747Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-5408deef3ec6127d.xml (deflated 84%) 2024-08-06T22:25:36.8482744Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-8c90e77e148b2752.xml (deflated 67%) 2024-08-06T22:25:36.8483756Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-0b59a7ae49ac25ca.xml (deflated 80%) 2024-08-06T22:25:36.8484781Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-94c9344efbd6dc1b.xml (deflated 29%) 2024-08-06T22:25:36.8485788Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-a9cee4f64f224215.xml (deflated 76%) 2024-08-06T22:25:36.8486932Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-6dd991447a836f65.xml (deflated 29%) 2024-08-06T22:25:36.8487948Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-732b5f356d37a9be.xml (deflated 34%) 2024-08-06T22:25:36.8488979Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-8f34bee56aa72227.xml (deflated 28%) 2024-08-06T22:25:36.8490002Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-63dc647abde41ca8.xml (deflated 48%) 2024-08-06T22:25:36.8491027Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-4f7ca3755400e7a7.xml (deflated 29%) 2024-08-06T22:25:36.8492043Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-9d2a11d8a6228a31.xml (deflated 60%) 2024-08-06T22:25:36.8493073Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-866e7c4cc74bc259.xml (deflated 28%) 2024-08-06T22:25:36.8494069Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-749240509acd8fc1.xml (deflated 36%) 2024-08-06T22:25:36.8495083Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-578e36301e10b5c3.xml (deflated 29%) 2024-08-06T22:25:36.8496096Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-be9f41af8783d60b.xml (deflated 72%) 2024-08-06T22:25:36.8497097Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-b8a86eb29b4318af.xml (deflated 29%) 2024-08-06T22:25:36.8498115Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-e59d3f9e832e6b94.xml (deflated 85%) 2024-08-06T22:25:36.8499131Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-566ac99c71cc7149.xml (deflated 29%) 2024-08-06T22:25:36.8500135Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-4731d6b6d01c750e.xml (deflated 51%) 2024-08-06T22:25:36.8501346Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-ff4928e6ad8174d5.xml (deflated 29%) 2024-08-06T22:25:36.8502377Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-05705466277ac539.xml (deflated 62%) 2024-08-06T22:25:36.8503386Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-cf09d745b78abb64.xml (deflated 29%) 2024-08-06T22:25:36.8507423Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-1e7140df4a722398.xml (deflated 88%) 2024-08-06T22:25:36.8508443Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-653e4887ebf10e8c.xml (deflated 28%) 2024-08-06T22:25:36.8521641Z adding: test/test-reports/python-pytest/test.run_test/test.run_test-595d30a92a33af94.xml (deflated 89%) 2024-08-06T22:25:36.8566219Z ##[group]Run # Remove any previous usage logs if they exist 2024-08-06T22:25:36.8566809Z # Remove any previous usage logs if they exist 2024-08-06T22:25:36.8567293Z rm -f logs-*.zip 2024-08-06T22:25:36.8567872Z # this workflow is also run in bazel build test, but we dont generate usage reports for it 2024-08-06T22:25:36.8568565Z # so check to see if the file exists first 2024-08-06T22:25:36.8569027Z if [ -f 'usage_log.txt' ]; then 2024-08-06T22:25:36.8569479Z  zip "logs-${FILE_SUFFIX}.zip" 'usage_log.txt' 2024-08-06T22:25:36.8569927Z fi 2024-08-06T22:25:36.8570270Z if ls test/**/*.log 1> /dev/null 2>&1; then 2024-08-06T22:25:36.8570759Z  zip -r "logs-${FILE_SUFFIX}.zip" test -i '*.log' 2024-08-06T22:25:36.8571208Z fi 2024-08-06T22:25:36.8577995Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T22:25:36.8578456Z env: 2024-08-06T22:25:36.8578746Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:36.8579172Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:36.8579843Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:36.8580660Z FILE_SUFFIX: test-default-1-5-amz2023.linux.4xlarge.nvidia.gpu_28428065757 2024-08-06T22:25:36.8581355Z ##[endgroup] 2024-08-06T22:25:36.8697814Z adding: usage_log.txt (deflated 92%) 2024-08-06T22:25:36.8976816Z adding: test/test-reports/test_native_mha_1.1_43f467f7830589b4_.log (deflated 94%) 2024-08-06T22:25:36.9048910Z adding: test/test-reports/test_nn_1.1_5430fad9bab7e294_.log (deflated 96%) 2024-08-06T22:25:36.9080206Z adding: test/test-reports/test_overrides_1.1_365907168bbc09ff_.log (deflated 93%) 2024-08-06T22:25:36.9095139Z adding: test/test-reports/test_cpp_api_parity_1.1_d92db0df232208df_.log (deflated 95%) 2024-08-06T22:25:36.9124042Z adding: test/test-reports/test_torch_1.1_38c76b710a46fc3c_.log (deflated 91%) 2024-08-06T22:25:36.9125950Z adding: test/test-reports/test_namedtuple_return_api_1.1_f9107d81858f3a37_.log (deflated 61%) 2024-08-06T22:25:36.9127882Z adding: test/test-reports/test_decomp_5.24_c3f5a8e4c029509f_.log (deflated 48%) 2024-08-06T22:25:36.9129663Z adding: test/test-reports/test_decomp_14.24_e42667637da2d357_.log (deflated 48%) 2024-08-06T22:25:36.9131424Z adding: test/test-reports/test_decomp_15.24_53521d8677d3baa5_.log (deflated 49%) 2024-08-06T22:25:36.9133183Z adding: test/test-reports/test_decomp_24.24_4ae095422df96e07_.log (deflated 48%) 2024-08-06T22:25:36.9135119Z adding: test/test-reports/test_model_exports_to_core_aten_1.1_92e0e483d1c466b3_.log (deflated 51%) 2024-08-06T22:25:36.9136742Z adding: test/test-reports/test_transformers_1.1_8dcd81afa21e5d8d_.log (deflated 50%) 2024-08-06T22:25:36.9137572Z adding: test/test-reports/functorch.test_ops_8.9_b9165706ceebc288_.log (deflated 49%) 2024-08-06T22:25:36.9138400Z adding: test/test-reports/test_proxy_tensor_1.1_e6667a18940f47c4_.log (deflated 49%) 2024-08-06T22:25:36.9139168Z adding: test/test-reports/test_foreach_1.1_0fcd8f2885d33bb7_.log (deflated 49%) 2024-08-06T22:25:36.9140147Z adding: test/test-reports/test_ops_gradients_1.4_a9ece66daa539b1c_.log (deflated 49%) 2024-08-06T22:25:36.9141072Z adding: test/test-reports/test_meta_5.5_ec05767ff76bc58d_.log (deflated 49%) 2024-08-06T22:25:36.9141817Z adding: test/test-reports/test_ops_2.11_dc32d920906cf18b_.log (deflated 49%) 2024-08-06T22:25:36.9142542Z adding: test/test-reports/test_ops_9.11_0a0a65c8b31e19fe_.log (deflated 49%) 2024-08-06T22:25:36.9143319Z adding: test/test-reports/test_module_tracker_1.1_e6469be478c2e430_.log (deflated 50%) 2024-08-06T22:25:36.9144138Z adding: test/test-reports/test_comparison_utils_1.1_ec0fbb2788e0faa4_.log (deflated 50%) 2024-08-06T22:25:36.9144983Z adding: test/test-reports/functorch.test_vmap_1.1_c8522dc184c01937_.log (deflated 50%) 2024-08-06T22:25:36.9145847Z adding: test/test-reports/inductor.test_inplacing_pass_1.1_8e394ec6337397f3_.log (stored 0%) 2024-08-06T22:25:36.9146692Z adding: test/test-reports/export.test_tools_1.1_e61989030b35495d_.log (deflated 50%) 2024-08-06T22:25:36.9174831Z adding: test/test-reports/cpp.test_api_1.1_1533c031bc1e263b_.log (deflated 93%) 2024-08-06T22:25:36.9176739Z adding: test/test-reports/inductor.test_torchinductor_dynamic_shapes_1.4_5bea70ed43c933aa_.log (deflated 55%) 2024-08-06T22:25:36.9177745Z adding: test/test-reports/torch_np.test_binary_ufuncs_1.1_982b23dcb1a41e4e_.log (deflated 51%) 2024-08-06T22:25:36.9206928Z adding: test/test-reports/cpp.test_tensorexpr_1.1_ea8e56300758860e_.log (deflated 94%) 2024-08-06T22:25:36.9209079Z adding: test/test-reports/dynamo.test_inline_inbuilt_nn_modules_1.1_c19a33c832a18dcf_.log (deflated 52%) 2024-08-06T22:25:36.9211294Z adding: test/test-reports/inductor.test_inductor_utils_1.1_6685839cf3a248f3_.log (deflated 51%) 2024-08-06T22:25:36.9213376Z adding: test/test-reports/torch_np.test_unary_ufuncs_1.1_ca25b089fa697dca_.log (deflated 51%) 2024-08-06T22:25:36.9215515Z adding: test/test-reports/inductor.test_extension_backend_1.1_92556c499ca43ed7_.log (deflated 51%) 2024-08-06T22:25:36.9217665Z adding: test/test-reports/export.test_export_nonstrict_1.1_6fbca26bceb97d5c_.log (deflated 60%) 2024-08-06T22:25:36.9219696Z adding: test/test-reports/cpp.test_tensorexpr_1.1_01cf64f8660069a7_.log (deflated 48%) 2024-08-06T22:25:36.9221013Z adding: test/test-reports/inductor.test_triton_extension_backend_1.1_00cac446b197b5d7_.log (deflated 52%) 2024-08-06T22:25:36.9221975Z adding: test/test-reports/inductor.test_triton_kernels_1.1_05e9541c9b0db076_.log (deflated 51%) 2024-08-06T22:25:36.9222863Z adding: test/test-reports/dynamo.test_base_output_1.1_5f87dc0ecad03c02_.log (deflated 51%) 2024-08-06T22:25:36.9223707Z adding: test/test-reports/export.test_serdes_1.1_154a92ee259f8afe_.log (deflated 59%) 2024-08-06T22:25:36.9224518Z adding: test/test-reports/dynamo.test_interop_1.1_d1ce67b3889c666f_.log (deflated 50%) 2024-08-06T22:25:36.9225398Z adding: test/test-reports/inductor.test_compile_worker_1.1_1ba16cb14fd7b18d_.log (deflated 51%) 2024-08-06T22:25:36.9226320Z adding: test/test-reports/dynamo.test_fx_passes_pre_grad_1.1_4ed62af637d74e6c_.log (deflated 51%) 2024-08-06T22:25:36.9227252Z adding: test/test-reports/inductor.test_flex_decoding_1.1_b9cd1c4c95d146cc_.log (deflated 51%) 2024-08-06T22:25:36.9228164Z adding: test/test-reports/dynamo.test_skip_non_tensor_1.1_8d8d7967a644ae3f_.log (deflated 51%) 2024-08-06T22:25:36.9229111Z adding: test/test-reports/inductor.test_move_constructors_to_cuda_1.1_9dcb182ea2d56995_.log (stored 0%) 2024-08-06T22:25:36.9230029Z adding: test/test-reports/test_matmul_cuda_1.1_28e8348588e14723_.log (deflated 49%) 2024-08-06T22:25:36.9230804Z adding: test/test-reports/dynamo.test_sdpa_1.1_e6e2d1c104b92330_.log (deflated 50%) 2024-08-06T22:25:36.9231619Z adding: test/test-reports/inductor.test_metrics_1.1_59af11765b8ea9e9_.log (stored 0%) 2024-08-06T22:25:36.9233315Z adding: test/test-reports/inductor.test_debug_trace_1.1_c75847ce10512bcf_.log (deflated 51%) 2024-08-06T22:25:36.9234143Z adding: test/test-reports/test_modules_1.2_fa9b04ae345eaf52_.log (deflated 48%) 2024-08-06T22:25:36.9235247Z adding: test/test-reports/inductor.test_cuda_cpp_wrapper_1.1_d07af4907c36874a_.log (stored 0%) 2024-08-06T22:25:36.9236098Z adding: test/test-reports/test_decomp_5.24_189b6118962869f4_.log (deflated 89%) 2024-08-06T22:25:36.9244861Z adding: test/test-reports/test_decomp_14.24_2c4b0a38c842b87b_.log (deflated 89%) 2024-08-06T22:25:36.9255761Z adding: test/test-reports/test_decomp_15.24_de5e9d44f2bd194a_.log (deflated 89%) 2024-08-06T22:25:36.9257235Z adding: test/test-reports/test_model_exports_to_core_aten_1.1_17753ab4128f7cce_.log (deflated 52%) 2024-08-06T22:25:36.9268098Z adding: test/test-reports/test_decomp_24.24_c4055e2d41561e4a_.log (deflated 89%) 2024-08-06T22:25:37.0768180Z adding: test/test-reports/test_transformers_1.1_95af2e2cb07b8232_.log (deflated 99%) 2024-08-06T22:25:37.0773550Z adding: test/test-reports/test_proxy_tensor_1.1_c073645f88a0b17f_.log (deflated 89%) 2024-08-06T22:25:37.0865939Z adding: test/test-reports/test_foreach_1.1_da04e0ec03200e2d_.log (deflated 95%) 2024-08-06T22:25:37.0901610Z adding: test/test-reports/functorch.test_ops_8.9_32afb2878ef66931_.log (deflated 92%) 2024-08-06T22:25:37.0944366Z adding: test/test-reports/test_ops_gradients_1.4_9a2dbcbd5d685f9d_.log (deflated 93%) 2024-08-06T22:25:37.1170272Z adding: test/test-reports/test_meta_5.5_2903666c02d5e4b6_.log (deflated 94%) 2024-08-06T22:25:37.1259687Z adding: test/test-reports/test_ops_2.11_a0911fee89d45ee0_.log (deflated 92%) 2024-08-06T22:25:37.1261438Z adding: test/test-reports/test_module_tracker_1.1_1ed99d90ca62e7cc_.log (deflated 56%) 2024-08-06T22:25:37.1262279Z adding: test/test-reports/test_comparison_utils_1.1_9801de73def40cda_.log (deflated 66%) 2024-08-06T22:25:37.1348934Z adding: test/test-reports/test_ops_9.11_8c2f83fcb8d6dbc6_.log (deflated 92%) 2024-08-06T22:25:37.1350423Z adding: test/test-reports/inductor.test_inplacing_pass_1.1_5b183ea65d3b7250_.log (stored 0%) 2024-08-06T22:25:37.1351283Z adding: test/test-reports/export.test_tools_1.1_6607c3bcb2b9bbdb_.log (deflated 63%) 2024-08-06T22:25:37.1415171Z adding: test/test-reports/functorch.test_vmap_1.1_d471fca716b413f4_.log (deflated 93%) 2024-08-06T22:25:37.1417032Z adding: test/test-reports/torch_np.test_binary_ufuncs_1.1_6f8c0ad317152332_.log (deflated 86%) 2024-08-06T22:25:37.1458780Z adding: test/test-reports/dynamo.test_inline_inbuilt_nn_modules_1.1_3b1c8a8197223013_.log (deflated 94%) 2024-08-06T22:25:37.1460565Z adding: test/test-reports/inductor.test_inductor_utils_1.1_629e1c9620efd0bf_.log (deflated 57%) 2024-08-06T22:25:37.1462165Z adding: test/test-reports/torch_np.test_unary_ufuncs_1.1_fd5722761f1a66f9_.log (deflated 86%) 2024-08-06T22:25:37.1463071Z adding: test/test-reports/inductor.test_extension_backend_1.1_6f606e5b2191a99d_.log (deflated 60%) 2024-08-06T22:25:37.1490575Z adding: test/test-reports/export.test_export_nonstrict_1.1_e932317b1a2734d9_.log (deflated 93%) 2024-08-06T22:25:37.1492009Z adding: test/test-reports/inductor.test_triton_extension_backend_1.1_846fcb8ddfbb5824_.log (deflated 55%) 2024-08-06T22:25:37.1497641Z adding: test/test-reports/inductor.test_triton_kernels_1.1_2aed89d432dfb97b_.log (deflated 94%) 2024-08-06T22:25:37.1499017Z adding: test/test-reports/dynamo.test_base_output_1.1_d8394e509c5348f9_.log (deflated 68%) 2024-08-06T22:25:37.1527577Z adding: test/test-reports/export.test_serdes_1.1_13f5b38669279a9b_.log (deflated 92%) 2024-08-06T22:25:37.1529549Z adding: test/test-reports/dynamo.test_interop_1.1_e7dd8e7c2ac12d8c_.log (deflated 61%) 2024-08-06T22:25:37.1530662Z adding: test/test-reports/inductor.test_compile_worker_1.1_20d1a87fa1171069_.log (deflated 57%) 2024-08-06T22:25:37.1531588Z adding: test/test-reports/dynamo.test_fx_passes_pre_grad_1.1_7bc28a1b2f7f494f_.log (deflated 53%) 2024-08-06T22:25:37.1534798Z adding: test/test-reports/inductor.test_flex_decoding_1.1_cc23c085ddf11236_.log (deflated 94%) 2024-08-06T22:25:37.1536340Z adding: test/test-reports/dynamo.test_skip_non_tensor_1.1_25dff3f0ff7c9d6e_.log (deflated 72%) 2024-08-06T22:25:37.1538801Z adding: test/test-reports/inductor.test_move_constructors_to_cuda_1.1_6012d3b3419e36c4_.log (stored 0%) 2024-08-06T22:25:37.1540149Z adding: test/test-reports/test_matmul_cuda_1.1_74baeb6da5259ad7_.log (deflated 87%) 2024-08-06T22:25:37.1540963Z adding: test/test-reports/dynamo.test_sdpa_1.1_bc8e6c4d8a4e0ef2_.log (deflated 60%) 2024-08-06T22:25:37.1541768Z adding: test/test-reports/inductor.test_metrics_1.1_a9ecdd0dd9f27f4e_.log (stored 0%) 2024-08-06T22:25:37.1542631Z adding: test/test-reports/inductor.test_debug_trace_1.1_8ab792b616f63b4e_.log (deflated 55%) 2024-08-06T22:25:37.1546829Z adding: test/test-reports/inductor.test_torchinductor_dynamic_shapes_1.4_cb3586c9672e681e_.log (deflated 91%) 2024-08-06T22:25:37.1547960Z adding: test/test-reports/inductor.test_cuda_cpp_wrapper_1.1_937d11bbc077ebd2_.log (stored 0%) 2024-08-06T22:25:37.1593155Z adding: test/test-reports/test_modules_1.2_8c1ccc7699308450_.log (deflated 94%) 2024-08-06T22:25:37.1594927Z adding: test/test-reports/cpp.Dict_test_1.1_b8a93d126e35bc15_.log (deflated 49%) 2024-08-06T22:25:37.1596775Z adding: test/test-reports/cpp.Dimname_test_1.1_159a6d4beede0f99_.log (deflated 49%) 2024-08-06T22:25:37.1598397Z adding: test/test-reports/cpp.NamedTensor_test_1.1_c2da3b1622e986d3_.log (deflated 48%) 2024-08-06T22:25:37.1599326Z adding: test/test-reports/cpp.apply_utils_test_1.1_6ce8fa5d18b43412_.log (deflated 49%) 2024-08-06T22:25:37.1601084Z adding: test/test-reports/cpp.atest_1.1_99622619a278fa84_.log (deflated 49%) 2024-08-06T22:25:37.1602864Z adding: test/test-reports/cpp.Dimname_test_1.1_6f9cae03abc80a0b_.log (deflated 60%) 2024-08-06T22:25:37.1605072Z adding: test/test-reports/cpp.NamedTensor_test_1.1_8d64a4fa755a2ae5_.log (deflated 72%) 2024-08-06T22:25:37.1606961Z adding: test/test-reports/cpp.Dict_test_1.1_b3dae9fe08b8f271_.log (deflated 85%) 2024-08-06T22:25:37.1608792Z adding: test/test-reports/cpp.apply_utils_test_1.1_dc73b066b157f97a_.log (deflated 66%) 2024-08-06T22:25:37.1610619Z adding: test/test-reports/cpp.atest_1.1_e3aa93d34d3c54a2_.log (deflated 75%) 2024-08-06T22:25:37.1612114Z adding: test/test-reports/cpp.cuda_generator_test_1.1_a341f51949e2695d_.log (deflated 49%) 2024-08-06T22:25:37.1614297Z adding: test/test-reports/cpp.cuda_generator_test_1.1_63920664cba4ff58_.log (deflated 75%) 2024-08-06T22:25:37.1616059Z adding: test/test-reports/cpp.cuda_half_test_1.1_8b27844259643beb_.log (deflated 48%) 2024-08-06T22:25:37.1616869Z adding: test/test-reports/cpp.cuda_half_test_1.1_7a878d97c4d54838_.log (deflated 49%) 2024-08-06T22:25:37.1617724Z adding: test/test-reports/cpp.cuda_vectorized_test_1.1_899f90a4ad4cc144_.log (deflated 49%) 2024-08-06T22:25:37.1618578Z adding: test/test-reports/cpp.cuda_vectorized_test_1.1_1c80927537e8ad2d_.log (deflated 56%) 2024-08-06T22:25:37.1619469Z adding: test/test-reports/cpp.cuda_distributions_test_1.1_2a7266db1843eeaf_.log (deflated 49%) 2024-08-06T22:25:37.1620389Z adding: test/test-reports/cpp.cuda_distributions_test_1.1_5a7d63b3b0a942b3_.log (deflated 64%) 2024-08-06T22:25:37.1621269Z adding: test/test-reports/cpp.cuda_optional_test_1.1_13f81acd68833150_.log (deflated 49%) 2024-08-06T22:25:37.1622401Z adding: test/test-reports/cpp.cuda_optional_test_1.1_1f16a7523e5d8b3d_.log (deflated 50%) 2024-08-06T22:25:37.1623266Z adding: test/test-reports/cpp.cuda_complex_test_1.1_d8c71e4d1b8456a5_.log (deflated 49%) 2024-08-06T22:25:37.1624110Z adding: test/test-reports/cpp.cuda_complex_test_1.1_0a521ef059b08c5a_.log (deflated 72%) 2024-08-06T22:25:37.1624958Z adding: test/test-reports/cpp.cuda_complex_math_test_1.1_9833f3c0d8743532_.log (deflated 49%) 2024-08-06T22:25:37.1625847Z adding: test/test-reports/cpp.cuda_complex_math_test_1.1_9f6b435f13a3abac_.log (deflated 83%) 2024-08-06T22:25:37.1626692Z adding: test/test-reports/cpp.cuda_cub_test_1.1_e3f50c3bd7708007_.log (deflated 49%) 2024-08-06T22:25:37.1627479Z adding: test/test-reports/cpp.cuda_cub_test_1.1_aefe592eb78e04d1_.log (deflated 57%) 2024-08-06T22:25:37.1628539Z adding: test/test-reports/cpp.cuda_atomic_ops_test_1.1_638a5c21685dd3c4_.log (deflated 49%) 2024-08-06T22:25:37.1629417Z adding: test/test-reports/cpp.cuda_atomic_ops_test_1.1_1e8239c3859eb2b3_.log (deflated 63%) 2024-08-06T22:25:37.1630217Z adding: test/test-reports/cpp.test_api_1.1_061a79b2722d930b_.log (deflated 48%) 2024-08-06T22:25:37.1669522Z ##[group]Run # Remove any previous debugging artifacts if they exist 2024-08-06T22:25:37.1670258Z # Remove any previous debugging artifacts if they exist 2024-08-06T22:25:37.1670748Z rm -f debug-*.zip 2024-08-06T22:25:37.1671115Z if [ -d 'test/debug' ]; then 2024-08-06T22:25:37.1671574Z  zip -r "debug-${FILE_SUFFIX}.zip" test/debug 2024-08-06T22:25:37.1671992Z fi 2024-08-06T22:25:37.1679653Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T22:25:37.1680134Z env: 2024-08-06T22:25:37.1680403Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:37.1680847Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:37.1681625Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:37.1682409Z FILE_SUFFIX: test-default-1-5-amz2023.linux.4xlarge.nvidia.gpu_28428065757 2024-08-06T22:25:37.1682986Z ##[endgroup] 2024-08-06T22:25:37.1791302Z ##[group]Run seemethere/upload-artifact-s3@v5 2024-08-06T22:25:37.1791721Z with: 2024-08-06T22:25:37.1792022Z s3-bucket: gha-artifacts 2024-08-06T22:25:37.1792439Z s3-prefix: pytorch/pytorch/10273124344/1/artifact 2024-08-06T22:25:37.1792872Z retention-days: 14 2024-08-06T22:25:37.1793213Z if-no-files-found: warn 2024-08-06T22:25:37.1793572Z path: test-jsons-*.zip 2024-08-06T22:25:37.1793893Z name: artifact 2024-08-06T22:25:37.1794199Z region: us-east-1 2024-08-06T22:25:37.1794506Z env: 2024-08-06T22:25:37.1794773Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:37.1795205Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:37.1795896Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:37.1796500Z ##[endgroup] 2024-08-06T22:25:37.6570758Z NOTE: s3-prefix specified, ignoring name parameter 2024-08-06T22:25:37.6572234Z With the provided path, there will be 1 file uploaded 2024-08-06T22:25:37.6573391Z Uploading to s3 prefix: pytorch/pytorch/10273124344/1/artifact 2024-08-06T22:25:37.6629158Z Starting upload of test-jsons-test-default-1-5-amz2023.linux.4xlarge.nvidia.gpu_28428065757.zip 2024-08-06T22:25:37.8122306Z Finished upload of test-jsons-test-default-1-5-amz2023.linux.4xlarge.nvidia.gpu_28428065757.zip 2024-08-06T22:25:37.8342254Z ##[group]Run seemethere/upload-artifact-s3@v5 2024-08-06T22:25:37.8342691Z with: 2024-08-06T22:25:37.8342981Z s3-bucket: gha-artifacts 2024-08-06T22:25:37.8343377Z s3-prefix: pytorch/pytorch/10273124344/1/artifact 2024-08-06T22:25:37.8343824Z retention-days: 14 2024-08-06T22:25:37.8344163Z if-no-files-found: error 2024-08-06T22:25:37.8344521Z path: test-reports-*.zip 2024-08-06T22:25:37.8344865Z name: artifact 2024-08-06T22:25:37.8345175Z region: us-east-1 2024-08-06T22:25:37.8345459Z env: 2024-08-06T22:25:37.8345739Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:37.8346175Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:37.8346837Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:37.8347442Z ##[endgroup] 2024-08-06T22:25:38.2844330Z NOTE: s3-prefix specified, ignoring name parameter 2024-08-06T22:25:38.2845000Z With the provided path, there will be 1 file uploaded 2024-08-06T22:25:38.2845564Z Uploading to s3 prefix: pytorch/pytorch/10273124344/1/artifact 2024-08-06T22:25:38.2897145Z Starting upload of test-reports-test-default-1-5-amz2023.linux.4xlarge.nvidia.gpu_28428065757.zip 2024-08-06T22:25:38.4846999Z Finished upload of test-reports-test-default-1-5-amz2023.linux.4xlarge.nvidia.gpu_28428065757.zip 2024-08-06T22:25:38.5046519Z ##[group]Run seemethere/upload-artifact-s3@v5 2024-08-06T22:25:38.5047122Z with: 2024-08-06T22:25:38.5047402Z s3-bucket: gha-artifacts 2024-08-06T22:25:38.5047994Z s3-prefix: pytorch/pytorch/10273124344/1/artifact 2024-08-06T22:25:38.5048469Z retention-days: 14 2024-08-06T22:25:38.5048803Z if-no-files-found: ignore 2024-08-06T22:25:38.5049161Z path: logs-*.zip 2024-08-06T22:25:38.5049471Z name: artifact 2024-08-06T22:25:38.5049758Z region: us-east-1 2024-08-06T22:25:38.5050059Z env: 2024-08-06T22:25:38.5050346Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:38.5050759Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:38.5051447Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:38.5052058Z ##[endgroup] 2024-08-06T22:25:38.9534350Z NOTE: s3-prefix specified, ignoring name parameter 2024-08-06T22:25:38.9535245Z With the provided path, there will be 1 file uploaded 2024-08-06T22:25:38.9535819Z Uploading to s3 prefix: pytorch/pytorch/10273124344/1/artifact 2024-08-06T22:25:38.9588153Z Starting upload of logs-test-default-1-5-amz2023.linux.4xlarge.nvidia.gpu_28428065757.zip 2024-08-06T22:25:39.1780373Z Finished upload of logs-test-default-1-5-amz2023.linux.4xlarge.nvidia.gpu_28428065757.zip 2024-08-06T22:25:39.1981055Z ##[group]Run seemethere/upload-artifact-s3@v5 2024-08-06T22:25:39.1981496Z with: 2024-08-06T22:25:39.1981780Z s3-bucket: gha-artifacts 2024-08-06T22:25:39.1982199Z s3-prefix: pytorch/pytorch/10273124344/1/artifact 2024-08-06T22:25:39.1982653Z retention-days: 14 2024-08-06T22:25:39.1982977Z if-no-files-found: ignore 2024-08-06T22:25:39.1983342Z path: debug-*.zip 2024-08-06T22:25:39.1983653Z name: artifact 2024-08-06T22:25:39.1983943Z region: us-east-1 2024-08-06T22:25:39.1984258Z env: 2024-08-06T22:25:39.1984544Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:39.1984961Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:39.1985649Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:39.1986275Z ##[endgroup] 2024-08-06T22:25:39.6379247Z No files were found with the provided path: debug-*.zip. No artifacts will be uploaded. 2024-08-06T22:25:39.6578121Z ##[group]Run # shellcheck disable=SC2156 2024-08-06T22:25:39.6578732Z # shellcheck disable=SC2156 2024-08-06T22:25:39.6579451Z find . -iname "core.[1-9]*" -exec docker exec "${DOCKER_CONTAINER_ID}" sh -c "gdb python {} -ex 'bt' -ex 'q'" \; 2024-08-06T22:25:39.6586704Z shell: /usr/bin/bash -e {0} 2024-08-06T22:25:39.6587071Z env: 2024-08-06T22:25:39.6587367Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:39.6587786Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:39.6588480Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:39.6589097Z ##[endgroup] 2024-08-06T22:25:39.9521332Z ##[group]Run pytorch/test-infra/.github/actions/teardown-linux@main 2024-08-06T22:25:39.9521913Z with: 2024-08-06T22:25:39.9522172Z env: 2024-08-06T22:25:39.9522469Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:39.9522904Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:39.9523574Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:39.9524196Z ##[endgroup] 2024-08-06T22:25:39.9570137Z ##[group]Run set -eou pipefail 2024-08-06T22:25:39.9570558Z set -eou pipefail 2024-08-06T22:25:39.9570893Z  2024-08-06T22:25:39.9571361Z echo "Holding runner for 2 hours until all ssh sessions have logged out" 2024-08-06T22:25:39.9571942Z for _ in $(seq 1440); do 2024-08-06T22:25:39.9572356Z  # Break if no ssh session exists anymore 2024-08-06T22:25:39.9572808Z  if [ "$(who)" = "" ]; then 2024-08-06T22:25:39.9573197Z  break 2024-08-06T22:25:39.9573552Z  fi 2024-08-06T22:25:39.9573844Z  echo "." 2024-08-06T22:25:39.9574168Z  sleep 5 2024-08-06T22:25:39.9574625Z done 2024-08-06T22:25:39.9581386Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T22:25:39.9581865Z env: 2024-08-06T22:25:39.9582153Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:39.9582564Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:39.9583252Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:39.9583868Z ##[endgroup] 2024-08-06T22:25:39.9612512Z Holding runner for 2 hours until all ssh sessions have logged out 2024-08-06T22:25:39.9695448Z ##[group]Run # ignore expansion of "docker ps -q" since it could be empty 2024-08-06T22:25:39.9696143Z # ignore expansion of "docker ps -q" since it could be empty 2024-08-06T22:25:39.9696669Z # shellcheck disable=SC2046 2024-08-06T22:25:39.9697115Z docker stop $(docker ps -q) || true 2024-08-06T22:25:39.9697562Z # Prune all of the docker images 2024-08-06T22:25:39.9697961Z docker system prune -af 2024-08-06T22:25:39.9704704Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T22:25:39.9705193Z env: 2024-08-06T22:25:39.9705524Z GIT_DEFAULT_BRANCH: main 2024-08-06T22:25:39.9705941Z GPU_FLAG: --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all 2024-08-06T22:25:39.9706638Z DOCKER_CONTAINER_ID: 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:39.9707248Z ##[endgroup] 2024-08-06T22:25:40.6006033Z 492035230bda 2024-08-06T22:25:41.1599786Z Deleted Containers: 2024-08-06T22:25:41.1600677Z 492035230bda220e745a1bc7447814e1d2c6b0f780765c2cb3f86f3622af9c2e 2024-08-06T22:25:41.1601381Z 2024-08-06T22:25:46.4002475Z Deleted Images: 2024-08-06T22:25:46.4003521Z untagged: 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9:02ec4fbd5adcb3fb91cf5ce431dec18b633de7d9 2024-08-06T22:25:46.4005693Z untagged: 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9@sha256:00f47b036f588ca5ef8866f8635fabba5a95cdf9ff1adae7d2a674ef1d4076e9 2024-08-06T22:25:46.4007067Z deleted: sha256:6ec36276acd88c9be8b44d856744037d399b35f4bb1703e637c27ae2b254c901 2024-08-06T22:25:46.4007873Z deleted: sha256:6fbc5fe2ebb0dc33846ab2ade7c5296a0a521e16f71c3b15b8a0c40a8fce5ed3 2024-08-06T22:25:46.4008993Z deleted: sha256:31fb5fadd0be2cc6e0d198fbc00e4f3c25925bc9c9ce79be23f02aeb56f6a55a 2024-08-06T22:25:46.4009802Z deleted: sha256:d1d89c6e648d792c08fdaae4fdad1273f93571a1b8d03c76f38f9e2be7cfe7f1 2024-08-06T22:25:46.4010606Z deleted: sha256:0f1adf9d1a1d4eeb62caa063c1090ea2f50246ed031d8d6627df2f5fe5963067 2024-08-06T22:25:46.4011384Z deleted: sha256:771e69c2306df070c5a944ea3043f81ce5890ce90de66fb6f16edfaf09912b35 2024-08-06T22:25:46.4012164Z deleted: sha256:288aa01dcd110ad9e2e3f48756d987529733a41225b6e8d4898f7386666beeae 2024-08-06T22:25:46.4013841Z deleted: sha256:d9d7c6c9bef79d8c5ce30d89b81820f3664c04fe17ae396b2a121e96fcfdeccf 2024-08-06T22:25:46.4014672Z deleted: sha256:1efa95271d86fd98beb626e63f4c67732daca36bf6e25d864735e43ef4c708f4 2024-08-06T22:25:46.4015464Z deleted: sha256:b1699f7cf1967593f94664569bb49ba431deaac0d1ceaf0d0584046a78ca3be0 2024-08-06T22:25:46.4016252Z deleted: sha256:b70eb1cb717d1c2ce90d5fa7f48e1af32d45f3735b33e1f944762686d8459aa6 2024-08-06T22:25:46.4017048Z deleted: sha256:6460a71f50d3c4bf7773fa072665e881e2a7e0e4e0bd45302e6a73c35ec03898 2024-08-06T22:25:46.4017830Z deleted: sha256:1c1787b4980844f6bfea1fe2386125ac286e9dfd80e253164a7f82b90f9c37bc 2024-08-06T22:25:46.4018618Z deleted: sha256:758fb95d0e1c4b4c78ee59fc4d66c06acfbc3a3ffa870e94200529367b8998e8 2024-08-06T22:25:46.4019416Z deleted: sha256:99beb19b8911fc7c49d248960a97d8ba4bbcbe8afc9fe3142e5af37f08c1c821 2024-08-06T22:25:46.4020193Z deleted: sha256:55596f5748c7f4d8e4d9fea0d1a6cda4627ccc442e652c8578169efe5992a382 2024-08-06T22:25:46.4020992Z deleted: sha256:c279b10b40b22a2fba8aeaad85e7705ffb1afa28e000323b35cc947649b7637d 2024-08-06T22:25:46.4021804Z deleted: sha256:8c882a49501c1d973eacd2ed8da81bce818403153d3fb4cdc84baf14307f9517 2024-08-06T22:25:46.4022704Z deleted: sha256:9fdbb157fcea7486d127e106ff114401c3102b9e9ed879a59247a296b3a908af 2024-08-06T22:25:46.4023493Z deleted: sha256:25b821c2e5d4106b0214b3aa4c88265bda5f90f045f012af912f0a3d2979f919 2024-08-06T22:25:46.4024288Z deleted: sha256:fada342ffd1729d0c38b7afe6767d6840580548d8dac7c62ddf61e24803ad66e 2024-08-06T22:25:46.4025066Z deleted: sha256:14d55c394ec592bb1417a21f696089e0549403bff381cda63d3e92ec80bca298 2024-08-06T22:25:46.4025869Z deleted: sha256:1032dedb70a12fc3c1cae4b10cdd7a07a9d20a64736533d640dd133d46f1ebfe 2024-08-06T22:25:46.4026659Z deleted: sha256:c0f5baad08f078d4210c254240524d9063c5e797564ada8a4bf39270e0fc1300 2024-08-06T22:25:46.4027437Z deleted: sha256:1a8bc02cd2c8f897e1dfd14920703ea75d172660c85026068b39136a1a25db51 2024-08-06T22:25:46.4028202Z deleted: sha256:b2221232f25b8d3a916bf8f74248542af7830e6198814ffbdbab43484bcb700b 2024-08-06T22:25:46.4028990Z deleted: sha256:14dc7064caf62d3d51b65e5175f9ae0e31af4fbed83413736f80d881f0fd742e 2024-08-06T22:25:46.4029786Z deleted: sha256:162ef080ee355a240e7e8fd6761113ad61586e1c883207ba4290c90abf208b54 2024-08-06T22:25:46.4030565Z deleted: sha256:56e3e897838040eb3fb86255ef1be2e3be6705d2e4a1ead67bbd400350ff6d13 2024-08-06T22:25:46.4031582Z deleted: sha256:a4f1392c3713fac843566ae891b9e41c28824e377af84190d33a4132cd4b268d 2024-08-06T22:25:46.4032379Z deleted: sha256:d1a8ab1bb8f3238d77db318f5161058057f84f7357afd2932e1d7edded9a2efb 2024-08-06T22:25:46.4033149Z deleted: sha256:7cab46d23f9fd91e62280b25d274d6c8a03be6d66030adf598944913d280c54f 2024-08-06T22:25:46.4033939Z deleted: sha256:764da7f97ae4318f5157e363c03c86c016b8d1b6d2c4758d203fa8f194753f57 2024-08-06T22:25:46.4034720Z deleted: sha256:373ab2f02049fb6ccab3f886c53934457751a14e5cb68b5e144b04bb7afeed87 2024-08-06T22:25:46.4035492Z deleted: sha256:e46d90cf4cc84661ac93ae9bcfbc675c0fab0582328e4595d075e4f259c4389a 2024-08-06T22:25:46.4036294Z deleted: sha256:1579b987297253e32b13fe0d160a95c567910203783be2efcc3a76650255e658 2024-08-06T22:25:46.4037074Z deleted: sha256:f8542342382d7443bc8be95e90ac743f70abdafcd9987e3785f58ad50509a145 2024-08-06T22:25:46.4037854Z deleted: sha256:57432242bf8090a0c1238a5f004be4da246f4488a2193be182e5cd82237509e4 2024-08-06T22:25:46.4038640Z deleted: sha256:e3c369d15f59fb5f34178d8dcb6bb4a6ff1d527e9baf3182444c24c55600bdf4 2024-08-06T22:25:46.4039505Z deleted: sha256:4215189c6756d59c8a217cc09447c312318d07c0b0d5c9bfb7e4bfb942f05cae 2024-08-06T22:25:46.4040271Z deleted: sha256:b787e6329788262404e4b8293a110727833b63e2e287ccd569a21cb3fb450388 2024-08-06T22:25:46.4041256Z deleted: sha256:69bb5dca4b9fded8d0c731b73bcbed0c3e9ce170c89e79796956a444b0d58c4c 2024-08-06T22:25:46.4042301Z deleted: sha256:8c945f56e57fb4fbfa3f2c74d6109a47db8df7644c4988d0c5791e4214bf30c9 2024-08-06T22:25:46.4043105Z deleted: sha256:fdd079e0e07e11d86ffc744f1036793f3db2aea378660f7489fbf50002439620 2024-08-06T22:25:46.4044301Z deleted: sha256:6a743a461982e1a1b73f29cd187e354f2119f3bd985e1da6ca0b802134cb91ba 2024-08-06T22:25:46.4045109Z deleted: sha256:7b3cbce1a91c69d6a499ef524311c14537930fc04f0ef4b1d73030216ac4a568 2024-08-06T22:25:46.4045895Z deleted: sha256:d7120733e426141c8a9e8f2c3596b12055cd5b1956d141ad640365cb11628a00 2024-08-06T22:25:46.4046670Z deleted: sha256:841ae666d028560ea37e6314ae5f80d72fb071fe7592b577dbbd8156c08cdda2 2024-08-06T22:25:46.4047450Z deleted: sha256:9d85a8c4ea0c0bc631a4d3e5472e8ed8e2312a4dc5a38f6e9890edeb37183186 2024-08-06T22:25:46.4048240Z deleted: sha256:606f6418680b5c53158271dfaca16c668f3b9dd25bb2b738f5621b6a56d08cdb 2024-08-06T22:25:46.4049047Z deleted: sha256:5ca4df537c3efc44dfde3bebb377ed490bc4d1cd5d6e5fa2fd9549dcb456c471 2024-08-06T22:25:46.4049850Z deleted: sha256:c6f3f6b7969e62b95fe926a1963e6628cb2f1b5388f0b12115564855e59423f1 2024-08-06T22:25:46.4050651Z deleted: sha256:ce178391f5f25e4ebe6ba8e84c84831fb31b8a9d4e82dbc270a73ddeccbf1c2c 2024-08-06T22:25:46.4051465Z deleted: sha256:d154e0a00edb13898d564f38c090096dfeb7a90beb0bd8addd57eca33f52151e 2024-08-06T22:25:46.4052245Z deleted: sha256:7226a8914ccef500b1d0327370cf9e3a66fbcb0b59a5f3b018981372152fc645 2024-08-06T22:25:46.4053032Z deleted: sha256:ce5fab2081603cd22d9f253d98828ddd60c0f8c44c5413d0615f4264f1cd6a7b 2024-08-06T22:25:46.4053902Z deleted: sha256:b3bc21b2cb6a9ca2b1dbbc412812581089c8ace8cc6b8d2b2767f0b3cbe8b99c 2024-08-06T22:25:46.4054684Z deleted: sha256:53e8f21eb1afc2c8550bef59d67810a5f79671d1d3d3924577f810321e66886b 2024-08-06T22:25:46.4055482Z deleted: sha256:f9770918a5cc46e8f6aa13da6903754fa89a6bcc9f6e11df1d61f9340e452cc8 2024-08-06T22:25:46.4056287Z deleted: sha256:bce2de65cb97e6a64f59fe7fc78644a2f3d62cf7769073cb59dcbb52009fc5d0 2024-08-06T22:25:46.4057072Z deleted: sha256:568d161c6888a8b0b09d7feb561f790a31d47c7ff7ca1c89c0ff4025f5b02e3a 2024-08-06T22:25:46.4057864Z deleted: sha256:2da22d4dfb814e7a3f4444f60464109149d35bc3e078e197f600bd9f6cbb9f6b 2024-08-06T22:25:46.4058646Z deleted: sha256:4af2daf4a1277af00005823106d88d704d5d0499eef553f01875c6f94380b2e3 2024-08-06T22:25:46.4059462Z deleted: sha256:e52d382a6cbfc1bb5122186d4b1401de79de002a9cc323ba5047dbf266f4d3a3 2024-08-06T22:25:46.4060241Z deleted: sha256:d10fe93dc5314d752e8882c5877457a6ae733a93a1e246cd6f79301de9325e9c 2024-08-06T22:25:46.4061037Z deleted: sha256:58d5ecb60ded3f99102b25d909a729386d92a44bf68b77ed3b49bf27d978e26a 2024-08-06T22:25:46.4061843Z deleted: sha256:dd522c9c6fdd98e9e4584578a66d9122b36ee8f856bba2140fcaaf908c7a68e8 2024-08-06T22:25:46.4062630Z deleted: sha256:3584fb50d1cbc80b57118953f1dc36ecb092b5986d56211f711b9161f403d66c 2024-08-06T22:25:46.4063427Z deleted: sha256:35a033279d5c9bd6b862bfa9331d8fcbf98bbaf778a778fabc882384db9204f8 2024-08-06T22:25:46.4064220Z deleted: sha256:ab9b00c496073d62e69e032178858d3d6d4c4bab9e87065d3785c6da57351d00 2024-08-06T22:25:46.4064984Z deleted: sha256:7ed947299e109c2459de6e240d86f049c30f93f2380659ce441d8737b8cd065f 2024-08-06T22:25:46.4065787Z deleted: sha256:410d5a7f7a9a1cb4551c106b9cc728a9dcff598f9be231a351ce6a0a33f81e64 2024-08-06T22:25:46.4066591Z deleted: sha256:5f0021bb56efa14bb93978c01513e2a1187ba30e69bdb0546d0ef39b30873f88 2024-08-06T22:25:46.4067368Z deleted: sha256:c1601aa97eb84151c14da3aeea351201bd99144d36d66b397f6555d80245d86d 2024-08-06T22:25:46.4068185Z deleted: sha256:72af05a89be22accdc1ca5d66dcbbb33993a9ef5997f849df1d4ba4c48049f25 2024-08-06T22:25:46.4068993Z deleted: sha256:38b90c9663dcaa2bc57a4dd3008298e7ea93e9535fa42312b4dc4246e7491af9 2024-08-06T22:25:46.4069861Z deleted: sha256:bda61b6cefb3ec8eeb74fef1ca1c7f9a5845fe5e8f07b8123323e897425d5c29 2024-08-06T22:25:46.4070672Z deleted: sha256:5a18e1aa877074529a84cbddf19f8d5403787823378ceae6b72fb62f78d43037 2024-08-06T22:25:46.4071466Z deleted: sha256:6c3e7df31590f02f10cb71fc4eb27653e9b428df2e6e5421a455b062bd2e39f9 2024-08-06T22:25:46.4071932Z 2024-08-06T22:25:46.4072114Z Total reclaimed space: 25.18GB 2024-08-06T22:25:46.4141335Z Post job cleanup. 2024-08-06T22:25:46.4201565Z Post job cleanup. 2024-08-06T22:25:46.5364328Z [command]/usr/bin/git version 2024-08-06T22:25:46.5420150Z git version 2.40.1 2024-08-06T22:25:46.5479895Z Temporarily overriding HOME='/home/ec2-user/actions-runner/_work/_temp/64954a04-7c54-4d9f-a24b-d005e44156ca' before making global git config changes 2024-08-06T22:25:46.5481453Z Adding repository directory to the temporary git global config as a safe directory 2024-08-06T22:25:46.5487357Z [command]/usr/bin/git config --global --add safe.directory /home/ec2-user/actions-runner/_work/pytorch/pytorch 2024-08-06T22:25:46.5533713Z [command]/usr/bin/git config --local --name-only --get-regexp core\.sshCommand 2024-08-06T22:25:46.5572511Z [command]/usr/bin/git submodule foreach --recursive sh -c "git config --local --name-only --get-regexp 'core\.sshCommand' && git config --local --unset-all 'core.sshCommand' || :" 2024-08-06T22:25:46.5929374Z Entering 'android/libs/fbjni' 2024-08-06T22:25:46.5992494Z Entering 'third_party/FP16' 2024-08-06T22:25:46.6056168Z Entering 'third_party/FXdiv' 2024-08-06T22:25:46.6119097Z Entering 'third_party/NNPACK' 2024-08-06T22:25:46.6182213Z Entering 'third_party/VulkanMemoryAllocator' 2024-08-06T22:25:46.6245413Z Entering 'third_party/XNNPACK' 2024-08-06T22:25:46.6329210Z Entering 'third_party/benchmark' 2024-08-06T22:25:46.6392469Z Entering 'third_party/cpp-httplib' 2024-08-06T22:25:46.6455541Z Entering 'third_party/cpuinfo' 2024-08-06T22:25:46.6519828Z Entering 'third_party/cudnn_frontend' 2024-08-06T22:25:46.6582534Z Entering 'third_party/cutlass' 2024-08-06T22:25:46.6654759Z Entering 'third_party/eigen' 2024-08-06T22:25:46.6720686Z Entering 'third_party/fbgemm' 2024-08-06T22:25:46.6784102Z Entering 'third_party/fbgemm/third_party/asmjit' 2024-08-06T22:25:46.6848013Z Entering 'third_party/fbgemm/third_party/cpuinfo' 2024-08-06T22:25:46.6912572Z Entering 'third_party/fbgemm/third_party/cutlass' 2024-08-06T22:25:46.6980387Z Entering 'third_party/fbgemm/third_party/googletest' 2024-08-06T22:25:46.7043386Z Entering 'third_party/fbgemm/third_party/hipify_torch' 2024-08-06T22:25:46.7106565Z Entering 'third_party/flatbuffers' 2024-08-06T22:25:46.7171419Z Entering 'third_party/fmt' 2024-08-06T22:25:46.7234091Z Entering 'third_party/foxi' 2024-08-06T22:25:46.7295449Z Entering 'third_party/gemmlowp/gemmlowp' 2024-08-06T22:25:46.7358852Z Entering 'third_party/gloo' 2024-08-06T22:25:46.7422003Z Entering 'third_party/googletest' 2024-08-06T22:25:46.7484986Z Entering 'third_party/ideep' 2024-08-06T22:25:46.7546630Z Entering 'third_party/ideep/mkl-dnn' 2024-08-06T22:25:46.7617848Z Entering 'third_party/ittapi' 2024-08-06T22:25:46.7679443Z Entering 'third_party/kineto' 2024-08-06T22:25:46.7742334Z Entering 'third_party/kineto/libkineto/third_party/dynolog' 2024-08-06T22:25:46.7803070Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/DCGM' 2024-08-06T22:25:46.7868111Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/cpr' 2024-08-06T22:25:46.7932778Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/fmt' 2024-08-06T22:25:46.7994792Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags' 2024-08-06T22:25:46.8056715Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags/doc' 2024-08-06T22:25:46.8122024Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/glog' 2024-08-06T22:25:46.8184035Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/googletest' 2024-08-06T22:25:46.8248065Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/json' 2024-08-06T22:25:46.8313191Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/pfs' 2024-08-06T22:25:46.8376069Z Entering 'third_party/kineto/libkineto/third_party/fmt' 2024-08-06T22:25:46.8438015Z Entering 'third_party/kineto/libkineto/third_party/googletest' 2024-08-06T22:25:46.8500768Z Entering 'third_party/mimalloc' 2024-08-06T22:25:46.8564169Z Entering 'third_party/nccl/nccl' 2024-08-06T22:25:46.8627321Z Entering 'third_party/nlohmann' 2024-08-06T22:25:46.8691877Z Entering 'third_party/onnx' 2024-08-06T22:25:46.8773965Z Entering 'third_party/onnx/third_party/benchmark' 2024-08-06T22:25:46.8837523Z Entering 'third_party/onnx/third_party/pybind11' 2024-08-06T22:25:46.8903378Z Entering 'third_party/opentelemetry-cpp' 2024-08-06T22:25:46.8968643Z Entering 'third_party/opentelemetry-cpp/third_party/benchmark' 2024-08-06T22:25:46.9031448Z Entering 'third_party/opentelemetry-cpp/third_party/googletest' 2024-08-06T22:25:46.9093506Z Entering 'third_party/opentelemetry-cpp/third_party/ms-gsl' 2024-08-06T22:25:46.9156371Z Entering 'third_party/opentelemetry-cpp/third_party/nlohmann-json' 2024-08-06T22:25:46.9221105Z Entering 'third_party/opentelemetry-cpp/third_party/opentelemetry-proto' 2024-08-06T22:25:46.9282104Z Entering 'third_party/opentelemetry-cpp/third_party/opentracing-cpp' 2024-08-06T22:25:46.9344903Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp' 2024-08-06T22:25:46.9406728Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/civetweb' 2024-08-06T22:25:46.9470338Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/googletest' 2024-08-06T22:25:46.9535848Z Entering 'third_party/opentelemetry-cpp/tools/vcpkg' 2024-08-06T22:25:46.9621774Z Entering 'third_party/pocketfft' 2024-08-06T22:25:46.9683571Z Entering 'third_party/protobuf' 2024-08-06T22:25:46.9750702Z Entering 'third_party/protobuf/third_party/benchmark' 2024-08-06T22:25:46.9814931Z Entering 'third_party/protobuf/third_party/googletest' 2024-08-06T22:25:46.9879266Z Entering 'third_party/psimd' 2024-08-06T22:25:46.9942221Z Entering 'third_party/pthreadpool' 2024-08-06T22:25:47.0004173Z Entering 'third_party/pybind11' 2024-08-06T22:25:47.0066639Z Entering 'third_party/python-peachpy' 2024-08-06T22:25:47.0129385Z Entering 'third_party/sleef' 2024-08-06T22:25:47.0191534Z Entering 'third_party/tensorpipe' 2024-08-06T22:25:47.0253886Z Entering 'third_party/tensorpipe/third_party/googletest' 2024-08-06T22:25:47.0315487Z Entering 'third_party/tensorpipe/third_party/libnop' 2024-08-06T22:25:47.0376560Z Entering 'third_party/tensorpipe/third_party/libuv' 2024-08-06T22:25:47.0441109Z Entering 'third_party/tensorpipe/third_party/pybind11' 2024-08-06T22:25:47.0501696Z Entering 'third_party/tensorpipe/third_party/pybind11/tools/clang' 2024-08-06T22:25:47.0587226Z [command]/usr/bin/git config --local --name-only --get-regexp http\.https\:\/\/github\.com\/\.extraheader 2024-08-06T22:25:47.0617724Z http.https://github.com/.extraheader 2024-08-06T22:25:47.0629306Z [command]/usr/bin/git config --local --unset-all http.https://github.com/.extraheader 2024-08-06T22:25:47.0669247Z [command]/usr/bin/git submodule foreach --recursive sh -c "git config --local --name-only --get-regexp 'http\.https\:\/\/github\.com\/\.extraheader' && git config --local --unset-all 'http.https://github.com/.extraheader' || :" 2024-08-06T22:25:47.1015130Z Entering 'android/libs/fbjni' 2024-08-06T22:25:47.1057037Z http.https://github.com/.extraheader 2024-08-06T22:25:47.1096606Z Entering 'third_party/FP16' 2024-08-06T22:25:47.1140116Z http.https://github.com/.extraheader 2024-08-06T22:25:47.1178625Z Entering 'third_party/FXdiv' 2024-08-06T22:25:47.1222408Z http.https://github.com/.extraheader 2024-08-06T22:25:47.1259935Z Entering 'third_party/NNPACK' 2024-08-06T22:25:47.1302797Z http.https://github.com/.extraheader 2024-08-06T22:25:47.1341916Z Entering 'third_party/VulkanMemoryAllocator' 2024-08-06T22:25:47.1384009Z http.https://github.com/.extraheader 2024-08-06T22:25:47.1423217Z Entering 'third_party/XNNPACK' 2024-08-06T22:25:47.1464659Z http.https://github.com/.extraheader 2024-08-06T22:25:47.1522977Z Entering 'third_party/benchmark' 2024-08-06T22:25:47.1564813Z http.https://github.com/.extraheader 2024-08-06T22:25:47.1603129Z Entering 'third_party/cpp-httplib' 2024-08-06T22:25:47.1646503Z http.https://github.com/.extraheader 2024-08-06T22:25:47.1683892Z Entering 'third_party/cpuinfo' 2024-08-06T22:25:47.1727043Z http.https://github.com/.extraheader 2024-08-06T22:25:47.1765411Z Entering 'third_party/cudnn_frontend' 2024-08-06T22:25:47.1809011Z http.https://github.com/.extraheader 2024-08-06T22:25:47.1846829Z Entering 'third_party/cutlass' 2024-08-06T22:25:47.1888995Z http.https://github.com/.extraheader 2024-08-06T22:25:47.1935835Z Entering 'third_party/eigen' 2024-08-06T22:25:47.1978052Z http.https://github.com/.extraheader 2024-08-06T22:25:47.2019323Z Entering 'third_party/fbgemm' 2024-08-06T22:25:47.2061707Z http.https://github.com/.extraheader 2024-08-06T22:25:47.2100579Z Entering 'third_party/fbgemm/third_party/asmjit' 2024-08-06T22:25:47.2143422Z http.https://github.com/.extraheader 2024-08-06T22:25:47.2181675Z Entering 'third_party/fbgemm/third_party/cpuinfo' 2024-08-06T22:25:47.2225482Z http.https://github.com/.extraheader 2024-08-06T22:25:47.2263421Z Entering 'third_party/fbgemm/third_party/cutlass' 2024-08-06T22:25:47.2305798Z http.https://github.com/.extraheader 2024-08-06T22:25:47.2352258Z Entering 'third_party/fbgemm/third_party/googletest' 2024-08-06T22:25:47.2394191Z http.https://github.com/.extraheader 2024-08-06T22:25:47.2432812Z Entering 'third_party/fbgemm/third_party/hipify_torch' 2024-08-06T22:25:47.2474939Z http.https://github.com/.extraheader 2024-08-06T22:25:47.2514924Z Entering 'third_party/flatbuffers' 2024-08-06T22:25:47.2557449Z http.https://github.com/.extraheader 2024-08-06T22:25:47.2598556Z Entering 'third_party/fmt' 2024-08-06T22:25:47.2641865Z http.https://github.com/.extraheader 2024-08-06T22:25:47.2679633Z Entering 'third_party/foxi' 2024-08-06T22:25:47.2723649Z http.https://github.com/.extraheader 2024-08-06T22:25:47.2761267Z Entering 'third_party/gemmlowp/gemmlowp' 2024-08-06T22:25:47.2803627Z http.https://github.com/.extraheader 2024-08-06T22:25:47.2841959Z Entering 'third_party/gloo' 2024-08-06T22:25:47.2884178Z http.https://github.com/.extraheader 2024-08-06T22:25:47.2922375Z Entering 'third_party/googletest' 2024-08-06T22:25:47.2964637Z http.https://github.com/.extraheader 2024-08-06T22:25:47.3002886Z Entering 'third_party/ideep' 2024-08-06T22:25:47.3046108Z http.https://github.com/.extraheader 2024-08-06T22:25:47.3082643Z Entering 'third_party/ideep/mkl-dnn' 2024-08-06T22:25:47.3125100Z http.https://github.com/.extraheader 2024-08-06T22:25:47.3171821Z Entering 'third_party/ittapi' 2024-08-06T22:25:47.3214992Z http.https://github.com/.extraheader 2024-08-06T22:25:47.3252491Z Entering 'third_party/kineto' 2024-08-06T22:25:47.3294704Z http.https://github.com/.extraheader 2024-08-06T22:25:47.3333678Z Entering 'third_party/kineto/libkineto/third_party/dynolog' 2024-08-06T22:25:47.3375162Z http.https://github.com/.extraheader 2024-08-06T22:25:47.3414655Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/DCGM' 2024-08-06T22:25:47.3456202Z http.https://github.com/.extraheader 2024-08-06T22:25:47.3496217Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/cpr' 2024-08-06T22:25:47.3538777Z http.https://github.com/.extraheader 2024-08-06T22:25:47.3577373Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/fmt' 2024-08-06T22:25:47.3619585Z http.https://github.com/.extraheader 2024-08-06T22:25:47.3658208Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags' 2024-08-06T22:25:47.3700246Z http.https://github.com/.extraheader 2024-08-06T22:25:47.3738974Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/gflags/doc' 2024-08-06T22:25:47.3780372Z http.https://github.com/.extraheader 2024-08-06T22:25:47.3821885Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/glog' 2024-08-06T22:25:47.3863306Z http.https://github.com/.extraheader 2024-08-06T22:25:47.3902526Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/googletest' 2024-08-06T22:25:47.3945865Z http.https://github.com/.extraheader 2024-08-06T22:25:47.3985810Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/json' 2024-08-06T22:25:47.4030560Z http.https://github.com/.extraheader 2024-08-06T22:25:47.4071024Z Entering 'third_party/kineto/libkineto/third_party/dynolog/third_party/pfs' 2024-08-06T22:25:47.4114636Z http.https://github.com/.extraheader 2024-08-06T22:25:47.4156970Z Entering 'third_party/kineto/libkineto/third_party/fmt' 2024-08-06T22:25:47.4199557Z http.https://github.com/.extraheader 2024-08-06T22:25:47.4239173Z Entering 'third_party/kineto/libkineto/third_party/googletest' 2024-08-06T22:25:47.4281103Z http.https://github.com/.extraheader 2024-08-06T22:25:47.4323265Z Entering 'third_party/mimalloc' 2024-08-06T22:25:47.4366298Z http.https://github.com/.extraheader 2024-08-06T22:25:47.4406370Z Entering 'third_party/nccl/nccl' 2024-08-06T22:25:47.4448753Z http.https://github.com/.extraheader 2024-08-06T22:25:47.4487643Z Entering 'third_party/nlohmann' 2024-08-06T22:25:47.4531193Z http.https://github.com/.extraheader 2024-08-06T22:25:47.4571333Z Entering 'third_party/onnx' 2024-08-06T22:25:47.4614389Z http.https://github.com/.extraheader 2024-08-06T22:25:47.4670451Z Entering 'third_party/onnx/third_party/benchmark' 2024-08-06T22:25:47.4714673Z http.https://github.com/.extraheader 2024-08-06T22:25:47.4752738Z Entering 'third_party/onnx/third_party/pybind11' 2024-08-06T22:25:47.4794662Z http.https://github.com/.extraheader 2024-08-06T22:25:47.4836972Z Entering 'third_party/opentelemetry-cpp' 2024-08-06T22:25:47.4879297Z http.https://github.com/.extraheader 2024-08-06T22:25:47.4920436Z Entering 'third_party/opentelemetry-cpp/third_party/benchmark' 2024-08-06T22:25:47.4961485Z http.https://github.com/.extraheader 2024-08-06T22:25:47.4999912Z Entering 'third_party/opentelemetry-cpp/third_party/googletest' 2024-08-06T22:25:47.5042507Z http.https://github.com/.extraheader 2024-08-06T22:25:47.5080750Z Entering 'third_party/opentelemetry-cpp/third_party/ms-gsl' 2024-08-06T22:25:47.5123445Z http.https://github.com/.extraheader 2024-08-06T22:25:47.5161187Z Entering 'third_party/opentelemetry-cpp/third_party/nlohmann-json' 2024-08-06T22:25:47.5203370Z http.https://github.com/.extraheader 2024-08-06T22:25:47.5243075Z Entering 'third_party/opentelemetry-cpp/third_party/opentelemetry-proto' 2024-08-06T22:25:47.5284829Z http.https://github.com/.extraheader 2024-08-06T22:25:47.5324253Z Entering 'third_party/opentelemetry-cpp/third_party/opentracing-cpp' 2024-08-06T22:25:47.5365837Z http.https://github.com/.extraheader 2024-08-06T22:25:47.5404357Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp' 2024-08-06T22:25:47.5446318Z http.https://github.com/.extraheader 2024-08-06T22:25:47.5483462Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/civetweb' 2024-08-06T22:25:47.5525781Z http.https://github.com/.extraheader 2024-08-06T22:25:47.5566596Z Entering 'third_party/opentelemetry-cpp/third_party/prometheus-cpp/3rdparty/googletest' 2024-08-06T22:25:47.5610179Z http.https://github.com/.extraheader 2024-08-06T22:25:47.5650571Z Entering 'third_party/opentelemetry-cpp/tools/vcpkg' 2024-08-06T22:25:47.5691923Z http.https://github.com/.extraheader 2024-08-06T22:25:47.5754510Z Entering 'third_party/pocketfft' 2024-08-06T22:25:47.5796512Z http.https://github.com/.extraheader 2024-08-06T22:25:47.5836219Z Entering 'third_party/protobuf' 2024-08-06T22:25:47.5879867Z http.https://github.com/.extraheader 2024-08-06T22:25:47.5924210Z Entering 'third_party/protobuf/third_party/benchmark' 2024-08-06T22:25:47.5967043Z http.https://github.com/.extraheader 2024-08-06T22:25:47.6005678Z Entering 'third_party/protobuf/third_party/googletest' 2024-08-06T22:25:47.6048073Z http.https://github.com/.extraheader 2024-08-06T22:25:47.6089618Z Entering 'third_party/psimd' 2024-08-06T22:25:47.6133648Z http.https://github.com/.extraheader 2024-08-06T22:25:47.6173134Z Entering 'third_party/pthreadpool' 2024-08-06T22:25:47.6218527Z http.https://github.com/.extraheader 2024-08-06T22:25:47.6256650Z Entering 'third_party/pybind11' 2024-08-06T22:25:47.6300067Z http.https://github.com/.extraheader 2024-08-06T22:25:47.6340560Z Entering 'third_party/python-peachpy' 2024-08-06T22:25:47.6383536Z http.https://github.com/.extraheader 2024-08-06T22:25:47.6422862Z Entering 'third_party/sleef' 2024-08-06T22:25:47.6464626Z http.https://github.com/.extraheader 2024-08-06T22:25:47.6503217Z Entering 'third_party/tensorpipe' 2024-08-06T22:25:47.6546351Z http.https://github.com/.extraheader 2024-08-06T22:25:47.6584442Z Entering 'third_party/tensorpipe/third_party/googletest' 2024-08-06T22:25:47.6628616Z http.https://github.com/.extraheader 2024-08-06T22:25:47.6666293Z Entering 'third_party/tensorpipe/third_party/libnop' 2024-08-06T22:25:47.6719349Z http.https://github.com/.extraheader 2024-08-06T22:25:47.6758294Z Entering 'third_party/tensorpipe/third_party/libuv' 2024-08-06T22:25:47.6800774Z http.https://github.com/.extraheader 2024-08-06T22:25:47.6840008Z Entering 'third_party/tensorpipe/third_party/pybind11' 2024-08-06T22:25:47.6882036Z http.https://github.com/.extraheader 2024-08-06T22:25:47.6921101Z Entering 'third_party/tensorpipe/third_party/pybind11/tools/clang' 2024-08-06T22:25:47.6963666Z http.https://github.com/.extraheader 2024-08-06T22:25:47.7111238Z A job completed hook has been configured by the self-hosted runner administrator 2024-08-06T22:25:47.7143411Z ##[group]Run '/home/ec2-user/runner-scripts/after_job.sh' 2024-08-06T22:25:47.7149477Z shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0} 2024-08-06T22:25:47.7149953Z ##[endgroup] 2024-08-06T22:25:55.8537885Z Cleaning up orphan processes